trainin_pipeline.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: SM_FRAMEWORK=tf.keras\n",
      "Segmentation Models: using `tf.keras` framework.\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "%env SM_FRAMEWORK=tf.keras\n",
    "import matplotlib.pyplot as plt\n",
    "from skimage.io import imread\n",
    "from skimage.transform import resize, rotate\n",
    "import os\n",
    "from os.path import join\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import backend as K\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tqdm.notebook import tqdm\n",
    "import albumentations as A\n",
    "import segmentation_models as sm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "\"\"\"AdamP for TensorFlow.\"\"\"\n",
    "from __future__ import absolute_import\n",
    "from __future__ import division\n",
    "from __future__ import print_function\n",
    "\n",
    "from tensorflow.python.framework import ops\n",
    "from tensorflow.python.keras import backend_config\n",
    "from tensorflow.python.keras.optimizer_v2 import optimizer_v2\n",
    "from tensorflow.python.ops import array_ops\n",
    "from tensorflow.python.ops import control_flow_ops\n",
    "from tensorflow.python.ops import math_ops\n",
    "from tensorflow.python.ops import state_ops\n",
    "\n",
    "\n",
    "class AdamP(optimizer_v2.OptimizerV2):\n",
    "    _HAS_AGGREGATE_GRAD = True\n",
    "\n",
    "    def __init__(self,\n",
    "                 learning_rate=0.001,\n",
    "                 beta_1=0.9,\n",
    "                 beta_2=0.999,\n",
    "                 epsilon=1e-8,\n",
    "                 weight_decay=0.0,\n",
    "                 delta=0.1, wd_ratio=0.1, nesterov=False,\n",
    "                 name='AdamP',\n",
    "                 **kwargs):\n",
    "\n",
    "        super(AdamP, self).__init__(name, **kwargs)\n",
    "        self._set_hyper('learning_rate', kwargs.get('lr', learning_rate))\n",
    "        self._set_hyper('beta_1', beta_1)\n",
    "        self._set_hyper('beta_2', beta_2)\n",
    "        self._set_hyper('delta', delta)\n",
    "        self._set_hyper('wd_ratio', wd_ratio)\n",
    "\n",
    "        self.epsilon = epsilon or backend_config.epsilon()\n",
    "        self.weight_decay = weight_decay\n",
    "        self.nesterov = nesterov\n",
    "\n",
    "    def _create_slots(self, var_list):\n",
    "        # Create slots for the first and second moments.\n",
    "        # Separate for-loops to respect the ordering of slot variables from v1.\n",
    "        for var in var_list:\n",
    "            self.add_slot(var, 'm')\n",
    "        for var in var_list:\n",
    "            self.add_slot(var, 'v')\n",
    "        for var in var_list:\n",
    "            self.add_slot(var, 'p')\n",
    "\n",
    "    def _prepare_local(self, var_device, var_dtype, apply_state):\n",
    "        super(AdamP, self)._prepare_local(var_device, var_dtype, apply_state)\n",
    "\n",
    "        local_step = math_ops.cast(self.iterations + 1, var_dtype)\n",
    "        beta_1_t = array_ops.identity(self._get_hyper('beta_1', var_dtype))\n",
    "        beta_2_t = array_ops.identity(self._get_hyper('beta_2', var_dtype))\n",
    "        beta_1_power = math_ops.pow(beta_1_t, local_step)\n",
    "        beta_2_power = math_ops.pow(beta_2_t, local_step)\n",
    "\n",
    "        lr = apply_state[(var_device, var_dtype)]['lr_t']\n",
    "        bias_correction1 = 1 - beta_1_power\n",
    "        bias_correction2 = 1 - beta_2_power\n",
    "\n",
    "        delta = array_ops.identity(self._get_hyper('delta', var_dtype))\n",
    "        wd_ratio = array_ops.identity(self._get_hyper('wd_ratio', var_dtype))\n",
    "\n",
    "        apply_state[(var_device, var_dtype)].update(\n",
    "            dict(\n",
    "                lr=lr,\n",
    "                epsilon=ops.convert_to_tensor_v2(self.epsilon, var_dtype),\n",
    "                weight_decay=ops.convert_to_tensor_v2(self.weight_decay, var_dtype),\n",
    "                beta_1_t=beta_1_t,\n",
    "                beta_1_power=beta_1_power,\n",
    "                one_minus_beta_1_t=1 - beta_1_t,\n",
    "                beta_2_t=beta_2_t,\n",
    "                beta_2_power=beta_2_power,\n",
    "                one_minus_beta_2_t=1 - beta_2_t,\n",
    "                bias_correction1=bias_correction1,\n",
    "                bias_correction2=bias_correction2,\n",
    "                delta=delta,\n",
    "                wd_ratio=wd_ratio))\n",
    "\n",
    "    def set_weights(self, weights):\n",
    "        params = self.weights\n",
    "        # If the weights are generated by Keras V1 optimizer, it includes vhats\n",
    "        # optimizer has 2x + 1 variables. Filter vhats out for compatibility.\n",
    "        num_vars = int((len(params) - 1) / 2)\n",
    "        if len(weights) == 3 * num_vars + 1:\n",
    "            weights = weights[:len(params)]\n",
    "        super(AdamP, self).set_weights(weights)\n",
    "\n",
    "    def _resource_apply_dense(self, grad, var, apply_state=None):\n",
    "        var_device, var_dtype = var.device, var.dtype.base_dtype\n",
    "        coefficients = ((apply_state or {}).get((var_device, var_dtype))\n",
    "                        or self._fallback_apply_state(var_device, var_dtype))\n",
    "\n",
    "        # m_t = beta1 * m + (1 - beta1) * g_t\n",
    "        m = self.get_slot(var, 'm')\n",
    "        m_scaled_g_values = grad * coefficients['one_minus_beta_1_t']\n",
    "        m_t = state_ops.assign(m, m * coefficients['beta_1_t'] + m_scaled_g_values, use_locking=self._use_locking)\n",
    "\n",
    "        # v_t = beta2 * v + (1 - beta2) * (g_t * g_t)\n",
    "        v = self.get_slot(var, 'v')\n",
    "        v_scaled_g_values = (grad * grad) * coefficients['one_minus_beta_2_t']\n",
    "        v_t = state_ops.assign(v, v * coefficients['beta_2_t'] + v_scaled_g_values, use_locking=self._use_locking)\n",
    "\n",
    "        denorm = (math_ops.sqrt(v_t) / math_ops.sqrt(coefficients['bias_correction2'])) + coefficients['epsilon']\n",
    "        step_size = coefficients['lr'] / coefficients['bias_correction1']\n",
    "\n",
    "        if self.nesterov:\n",
    "            perturb = (coefficients['beta_1_t'] * m_t + coefficients['one_minus_beta_1_t'] * grad) / denorm\n",
    "        else:\n",
    "            perturb = m_t / denorm\n",
    "\n",
    "        # Projection\n",
    "        wd_ratio = 1\n",
    "        if len(var.shape) > 1:\n",
    "            perturb, wd_ratio = self._projection(var, grad, perturb, coefficients['delta'], coefficients['wd_ratio'], coefficients['epsilon'])\n",
    "\n",
    "        # Weight decay\n",
    "\n",
    "        if self.weight_decay > 0:\n",
    "            var = state_ops.assign(var, var * (1 - coefficients['lr'] * coefficients['weight_decay'] * wd_ratio), use_locking=self._use_locking)\n",
    "\n",
    "        var_update = state_ops.assign_sub(var, step_size * perturb, use_locking=self._use_locking)\n",
    "\n",
    "        return control_flow_ops.group(*[var_update, m_t, v_t])\n",
    "\n",
    "\n",
    "    def _resource_apply_sparse(self, grad, var, indices, apply_state=None):\n",
    "\n",
    "        var_device, var_dtype = var.device, var.dtype.base_dtype\n",
    "        coefficients = ((apply_state or {}).get((var_device, var_dtype))\n",
    "                        or self._fallback_apply_state(var_device, var_dtype))\n",
    "        \"\"\"\n",
    "        Adam\n",
    "        \"\"\"\n",
    "        # m_t = beta1 * m + (1 - beta1) * g_t\n",
    "        m = self.get_slot(var, 'm')\n",
    "        m_scaled_g_values = grad * coefficients['one_minus_beta_1_t']\n",
    "        m_t = state_ops.assign(m, m * coefficients['beta_1_t'],\n",
    "                               use_locking=self._use_locking)\n",
    "        with ops.control_dependencies([m_t]):\n",
    "            m_t = self._resource_scatter_add(m, indices, m_scaled_g_values)\n",
    "\n",
    "        # v_t = beta2 * v + (1 - beta2) * (g_t * g_t)\n",
    "        v = self.get_slot(var, 'v')\n",
    "        v_scaled_g_values = (grad * grad) * coefficients['one_minus_beta_2_t']\n",
    "        v_t = state_ops.assign(v, v * coefficients['beta_2_t'],\n",
    "                               use_locking=self._use_locking)\n",
    "        with ops.control_dependencies([v_t]):\n",
    "            v_t = self._resource_scatter_add(v, indices, v_scaled_g_values)\n",
    "\n",
    "        denorm = (math_ops.sqrt(v_t) / math_ops.sqrt(coefficients['bias_correction2'])) + coefficients['epsilon']\n",
    "        step_size = coefficients['lr'] / coefficients['bias_correction1']\n",
    "\n",
    "        if self.nesterov:\n",
    "            p_scaled_g_values = grad * coefficients['one_minus_beta_1_t']\n",
    "            perturb = m_t * coefficients['beta_1_t']\n",
    "            perturb = self._resource_scatter_add(perturb, indices, p_scaled_g_values) / denorm\n",
    "\n",
    "        else:\n",
    "            perturb = m_t / denorm\n",
    "\n",
    "        # Projection\n",
    "        wd_ratio = 1\n",
    "        if len(var.shape) > 1:\n",
    "            perturb, wd_ratio = self._projection(var, grad, perturb, coefficients['delta'], coefficients['wd_ratio'], coefficients['epsilon'])\n",
    "\n",
    "        # Weight decay\n",
    "        if self.weight_decay > 0:\n",
    "            var = state_ops.assign(var, var * (1 - coefficients['lr'] * coefficients['weight_decay'] * wd_ratio), use_locking=self._use_locking)\n",
    "\n",
    "        var_update = state_ops.assign_sub(var, step_size * perturb, use_locking=self._use_locking)\n",
    "\n",
    "        return control_flow_ops.group(*[var_update, m_t, v_t])\n",
    "\n",
    "    def _channel_view(self, x):\n",
    "        return array_ops.reshape(x, shape=[x.shape[0], -1])\n",
    "\n",
    "    def _layer_view(self, x):\n",
    "        return array_ops.reshape(x, shape=[1, -1])\n",
    "\n",
    "    def _cosine_similarity(self, x, y, eps, view_func):\n",
    "        x = view_func(x)\n",
    "        y = view_func(y)\n",
    "\n",
    "        x_norm = math_ops.euclidean_norm(x, axis=-1) + eps\n",
    "        y_norm = math_ops.euclidean_norm(y, axis=-1) + eps\n",
    "        dot = math_ops.reduce_sum(x * y, axis=-1)\n",
    "\n",
    "        return math_ops.abs(dot) / x_norm / y_norm\n",
    "\n",
    "    def _projection(self, var, grad, perturb, delta, wd_ratio, eps):\n",
    "        # channel_view\n",
    "        cosine_sim = self._cosine_similarity(grad, var, eps, self._channel_view)\n",
    "        cosine_max = math_ops.reduce_max(cosine_sim)\n",
    "        compare_val = delta / math_ops.sqrt(math_ops.cast(self._channel_view(var).shape[-1], dtype=delta.dtype))\n",
    "\n",
    "        perturb, wd = control_flow_ops.cond(pred=cosine_max < compare_val,\n",
    "                                            true_fn=lambda : self.channel_true_fn(var, perturb, wd_ratio, eps),\n",
    "                                            false_fn=lambda : self.channel_false_fn(var, grad, perturb, delta, wd_ratio, eps))\n",
    "\n",
    "        return perturb, wd\n",
    "\n",
    "    def channel_true_fn(self, var, perturb, wd_ratio, eps):\n",
    "        expand_size = [-1] + [1] * (len(var.shape) - 1)\n",
    "        var_n = var / (array_ops.reshape(math_ops.euclidean_norm(self._channel_view(var), axis=-1), shape=expand_size) + eps)\n",
    "        perturb -= var_n * array_ops.reshape(math_ops.reduce_sum(self._channel_view(var_n * perturb), axis=-1), shape=expand_size)\n",
    "        wd = wd_ratio\n",
    "\n",
    "        return perturb, wd\n",
    "\n",
    "    def channel_false_fn(self, var, grad, perturb, delta, wd_ratio, eps):\n",
    "        cosine_sim = self._cosine_similarity(grad, var, eps, self._layer_view)\n",
    "        cosine_max = math_ops.reduce_max(cosine_sim)\n",
    "        compare_val = delta / math_ops.sqrt(math_ops.cast(self._layer_view(var).shape[-1], dtype=delta.dtype))\n",
    "\n",
    "        perturb, wd = control_flow_ops.cond(cosine_max < compare_val,\n",
    "                                              true_fn=lambda : self.layer_true_fn(var, perturb, wd_ratio, eps),\n",
    "                                              false_fn=lambda : self.identity_fn(perturb))\n",
    "\n",
    "        return perturb, wd\n",
    "\n",
    "    def layer_true_fn(self, var, perturb, wd_ratio, eps):\n",
    "        expand_size = [-1] + [1] * (len(var.shape) - 1)\n",
    "        var_n = var / (array_ops.reshape(math_ops.euclidean_norm(self._layer_view(var), axis=-1), shape=expand_size) + eps)\n",
    "        perturb -= var_n * array_ops.reshape(math_ops.reduce_sum(self._layer_view(var_n * perturb), axis=-1), shape=expand_size)\n",
    "        wd = wd_ratio\n",
    "\n",
    "        return perturb, wd\n",
    "\n",
    "    def identity_fn(self, perturb):\n",
    "        wd = 1.0\n",
    "\n",
    "        return perturb, wd\n",
    "\n",
    "    def get_config(self):\n",
    "        config = super(AdamP, self).get_config()\n",
    "        config.update({\n",
    "            'learning_rate': self._serialize_hyperparameter('learning_rate'),\n",
    "            'beta_1': self._serialize_hyperparameter('beta_1'),\n",
    "            'beta_2': self._serialize_hyperparameter('beta_2'),\n",
    "            'delta': self._serialize_hyperparameter('delta'),\n",
    "            'wd_ratio': self._serialize_hyperparameter('wd_ratio'),\n",
    "            'epsilon': self.epsilon,\n",
    "            'weight_decay': self.weight_decay,\n",
    "            'nesterov': self.nesterov\n",
    "        })\n",
    "        return config"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def visualize(image, mask, original_image=None, original_mask=None):\n",
    "    fontsize = 18\n",
    "    \n",
    "    if original_image is None and original_mask is None:\n",
    "        f, ax = plt.subplots(2, 1, figsize=(8, 8))\n",
    "\n",
    "        ax[0].imshow(image)\n",
    "        ax[1].imshow(mask)\n",
    "    else:\n",
    "        f, ax = plt.subplots(3, 1, figsize=(8, 8))\n",
    "\n",
    "        ax[0].imshow(image)\n",
    "        ax[0].set_title('GT', fontsize=fontsize)\n",
    "        \n",
    "        ax[1].imshow(mask)\n",
    "        ax[1].set_title('model mask', fontsize=fontsize)\n",
    "        \n",
    "        ax[2].imshow(original_image)\n",
    "        ax[2].set_title('Image', fontsize=fontsize)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_path = 'public_tests/00_test_val_input/train/images/'\n",
    "gt_path = 'public_tests/00_test_val_input/train/gt/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def read_and_resize_data(data_path, gt_path):\n",
    "    #TAKE = 15\n",
    "    size = 224\n",
    "    folders_data = np.array(sorted(os.listdir(data_path)))\n",
    "    folders_gt = np.array(sorted(os.listdir(gt_path)))\n",
    "    assert np.all(folders_gt == folders_data)\n",
    "    folders = folders_data\n",
    "    images_data = []\n",
    "    images_gt = []\n",
    "    for folder in tqdm(folders, desc='reading data folders...'):\n",
    "        files_data = np.array(sorted(os.listdir(data_path + folder)))\n",
    "        files_gt = np.array(sorted(os.listdir(gt_path + folder)))\n",
    "        files_data = [file[:-4] for file in files_data]\n",
    "        files_gt = [file[:-4] for file in files_gt]\n",
    "        #print(files_gt, files_data)\n",
    "        assert np.all(files_gt == files_data)\n",
    "        files = files_data\n",
    "        for file in files:\n",
    "            impath = data_path + folder + '/' + file + '.jpg'\n",
    "            gtpath = gt_path + folder + '/' + file + '.png'\n",
    "            image = imread(impath)\n",
    "            gt = imread(gtpath)\n",
    "            h, w = image.shape[:2]\n",
    "            images_data.append(resize(image, (size, size, 3)).astype('float16'))\n",
    "            positives = (resize(gt, (size, size, 1))).astype('float16')\n",
    "            negatives = (1 - resize(gt, (size, size, 1))).astype('float16')\n",
    "            gt = np.stack((negatives, positives), axis=2).reshape((size, size, 2))\n",
    "            images_gt.append(gt)\n",
    "    X = np.array(images_data)\n",
    "    Y = np.array(images_gt)\n",
    "    return X, Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4e3da90e22d44850a9d602597ff3a7ac",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='reading data folders...', max=200.0, style=ProgressStyle(…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "((8382, 224, 224, 3), (8382, 224, 224, 2))"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X, Y = read_and_resize_data(data_path, gt_path)\n",
    "X.shape, Y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.000e+00, 5.960e-08, 1.788e-07, ..., 9.990e-01, 9.995e-01,\n",
       "       1.000e+00], dtype=float16)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(Y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(224, 224, 3)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X[2].shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Generator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generator(x_train, y_train, batch_size=128):\n",
    "    cur = 0\n",
    "    indices = np.arange(x_train.shape[0])\n",
    "    np.random.shuffle(indices)\n",
    "    x_train = x_train[indices]\n",
    "    y_train = y_train[indices]\n",
    "    while True:\n",
    "        batchx = np.zeros((batch_size, 224, 224, 3))\n",
    "        batchy = np.zeros((batch_size, 224, 224, 2))\n",
    "        for i in range(batch_size):\n",
    "            transform = A.Compose([\n",
    "                    A.Flip(),\n",
    "                    A.RandomRotate90(),\n",
    "                    A.Transpose(),\n",
    "                    A.OneOf([\n",
    "                        A.RandomRain(),\n",
    "                        A.RandomFog(),\n",
    "                    ], p=0.2),\n",
    "                    A.GaussNoise(var_limit=(0.0005, 0.005)),\n",
    "                    #A.Blur(blur_limit=1.5, p=0.1),\n",
    "                    A.ShiftScaleRotate(shift_limit=0.0625, scale_limit=0.2, rotate_limit=40, p=0.8),\n",
    "                    A.OneOf([\n",
    "                        A.ElasticTransform(p=1, alpha=120, sigma=120 * 0.05, alpha_affine=120 * 0.03),\n",
    "                        A.OpticalDistortion(p=0.3),\n",
    "                        A.GridDistortion(p=.1),\n",
    "                        A.IAAPiecewiseAffine(p=0.3),\n",
    "                    ], p=0.2),\n",
    "                    #A.RandomBrightnessContrast(),            \n",
    "                ])\n",
    "            t = transform(image=x_train[cur].astype('float32'), mask=y_train[cur].astype('float32'))\n",
    "            batchx[i] = np.array(t['image'])\n",
    "            batchy[i] = np.array(t['mask'])\n",
    "            cur += 1\n",
    "            if cur == x_train.shape[0]:\n",
    "                batchx = batchx[:i+1, :]\n",
    "                batchy = batchy[:i+1, :]\n",
    "                indices = np.arange(x_train.shape[0])\n",
    "                np.random.shuffle(indices)\n",
    "                x_train = x_train[indices]\n",
    "                y_train = y_train[indices]\n",
    "                cur = 0\n",
    "                break\n",
    "        yield batchx.astype('float32'), batchy.astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9397244 0.0\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for x, y in generator(X, Y):\n",
    "    visualize(x[0], y[0][:, :, :1])\n",
    "    print(np.max(x[0]), np.min(x[0]))\n",
    "    break"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"functional_3\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "data (InputLayer)               [(None, 224, 224, 3) 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "bn_data (BatchNormalization)    (None, 224, 224, 3)  9           data[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d (ZeroPadding2D)  (None, 230, 230, 3)  0           bn_data[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv0 (Conv2D)                  (None, 112, 112, 64) 9408        zero_padding2d[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "bn0 (BatchNormalization)        (None, 112, 112, 64) 256         conv0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "relu0 (Activation)              (None, 112, 112, 64) 0           bn0[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_1 (ZeroPadding2D (None, 114, 114, 64) 0           relu0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "pooling0 (MaxPooling2D)         (None, 56, 56, 64)   0           zero_padding2d_1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_bn1 (BatchNormaliz (None, 56, 56, 64)   256         pooling0[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_2 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_3 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_3[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_sc (Conv2D)        (None, 56, 56, 64)   4096        stage1_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add (Add)                       (None, 56, 56, 64)   0           stage1_unit1_conv2[0][0]         \n",
      "                                                                 stage1_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_4 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_4[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_5 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_5[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "add_1 (Add)                     (None, 56, 56, 64)   0           stage1_unit2_conv2[0][0]         \n",
      "                                                                 add[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_6 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_6[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_7 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_7[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "add_2 (Add)                     (None, 56, 56, 64)   0           stage1_unit3_conv2[0][0]         \n",
      "                                                                 add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_relu1 (Activation) (None, 56, 56, 64)   0           stage2_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_8 (ZeroPadding2D (None, 58, 58, 64)   0           stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_conv1 (Conv2D)     (None, 28, 28, 128)  73728       zero_padding2d_8[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_9 (ZeroPadding2D (None, 30, 30, 128)  0           stage2_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_9[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_sc (Conv2D)        (None, 28, 28, 128)  8192        stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_3 (Add)                     (None, 28, 28, 128)  0           stage2_unit1_conv2[0][0]         \n",
      "                                                                 stage2_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_10 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_10[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_11 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_11[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_4 (Add)                     (None, 28, 28, 128)  0           stage2_unit2_conv2[0][0]         \n",
      "                                                                 add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_12 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_12[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_13 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_13[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_5 (Add)                     (None, 28, 28, 128)  0           stage2_unit3_conv2[0][0]         \n",
      "                                                                 add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit4_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_14 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit4_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_14[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit4_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit4_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_15 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit4_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_15[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_6 (Add)                     (None, 28, 28, 128)  0           stage2_unit4_conv2[0][0]         \n",
      "                                                                 add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_6[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_relu1 (Activation) (None, 28, 28, 128)  0           stage3_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_16 (ZeroPadding2 (None, 30, 30, 128)  0           stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_conv1 (Conv2D)     (None, 14, 14, 256)  294912      zero_padding2d_16[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_17 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_17[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_sc (Conv2D)        (None, 14, 14, 256)  32768       stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_7 (Add)                     (None, 14, 14, 256)  0           stage3_unit1_conv2[0][0]         \n",
      "                                                                 stage3_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_18 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_18[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_19 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_19[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_8 (Add)                     (None, 14, 14, 256)  0           stage3_unit2_conv2[0][0]         \n",
      "                                                                 add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_20 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_20[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_21 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_21[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_9 (Add)                     (None, 14, 14, 256)  0           stage3_unit3_conv2[0][0]         \n",
      "                                                                 add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit4_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_22 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit4_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_22[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit4_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit4_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_23 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit4_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_23[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_10 (Add)                    (None, 14, 14, 256)  0           stage3_unit4_conv2[0][0]         \n",
      "                                                                 add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit5_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_24 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit5_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_24[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit5_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit5_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_25 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit5_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_25[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_11 (Add)                    (None, 14, 14, 256)  0           stage3_unit5_conv2[0][0]         \n",
      "                                                                 add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit6_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_26 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit6_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_26[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit6_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit6_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_27 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit6_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_27[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_12 (Add)                    (None, 14, 14, 256)  0           stage3_unit6_conv2[0][0]         \n",
      "                                                                 add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_12[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_relu1 (Activation) (None, 14, 14, 256)  0           stage4_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_28 (ZeroPadding2 (None, 16, 16, 256)  0           stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_conv1 (Conv2D)     (None, 7, 7, 512)    1179648     zero_padding2d_28[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_29 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_29[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_sc (Conv2D)        (None, 7, 7, 512)    131072      stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_13 (Add)                    (None, 7, 7, 512)    0           stage4_unit1_conv2[0][0]         \n",
      "                                                                 stage4_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_bn1 (BatchNormaliz (None, 7, 7, 512)    2048        add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_relu1 (Activation) (None, 7, 7, 512)    0           stage4_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_30 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_conv1 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_30[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_31 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_31[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_14 (Add)                    (None, 7, 7, 512)    0           stage4_unit2_conv2[0][0]         \n",
      "                                                                 add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_bn1 (BatchNormaliz (None, 7, 7, 512)    2048        add_14[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_relu1 (Activation) (None, 7, 7, 512)    0           stage4_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_32 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_conv1 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_32[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_33 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_33[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_15 (Add)                    (None, 7, 7, 512)    0           stage4_unit3_conv2[0][0]         \n",
      "                                                                 add_14[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "bn1 (BatchNormalization)        (None, 7, 7, 512)    2048        add_15[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "relu1 (Activation)              (None, 7, 7, 512)    0           bn1[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0_upsampling (UpSa (None, 14, 14, 512)  0           relu1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0_concat (Concaten (None, 14, 14, 768)  0           decoder_stage0_upsampling[0][0]  \n",
      "                                                                 stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_conv (Conv2D)   (None, 14, 14, 256)  1769472     decoder_stage0_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_bn (BatchNormal (None, 14, 14, 256)  1024        decoder_stage0a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_relu (Activatio (None, 14, 14, 256)  0           decoder_stage0a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_conv (Conv2D)   (None, 14, 14, 256)  589824      decoder_stage0a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_bn (BatchNormal (None, 14, 14, 256)  1024        decoder_stage0b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_relu (Activatio (None, 14, 14, 256)  0           decoder_stage0b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1_upsampling (UpSa (None, 28, 28, 256)  0           decoder_stage0b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1_concat (Concaten (None, 28, 28, 384)  0           decoder_stage1_upsampling[0][0]  \n",
      "                                                                 stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_conv (Conv2D)   (None, 28, 28, 128)  442368      decoder_stage1_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_bn (BatchNormal (None, 28, 28, 128)  512         decoder_stage1a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_relu (Activatio (None, 28, 28, 128)  0           decoder_stage1a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_conv (Conv2D)   (None, 28, 28, 128)  147456      decoder_stage1a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_bn (BatchNormal (None, 28, 28, 128)  512         decoder_stage1b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_relu (Activatio (None, 28, 28, 128)  0           decoder_stage1b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2_upsampling (UpSa (None, 56, 56, 128)  0           decoder_stage1b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2_concat (Concaten (None, 56, 56, 192)  0           decoder_stage2_upsampling[0][0]  \n",
      "                                                                 stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_conv (Conv2D)   (None, 56, 56, 64)   110592      decoder_stage2_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_bn (BatchNormal (None, 56, 56, 64)   256         decoder_stage2a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_relu (Activatio (None, 56, 56, 64)   0           decoder_stage2a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_conv (Conv2D)   (None, 56, 56, 64)   36864       decoder_stage2a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_bn (BatchNormal (None, 56, 56, 64)   256         decoder_stage2b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_relu (Activatio (None, 56, 56, 64)   0           decoder_stage2b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3_upsampling (UpSa (None, 112, 112, 64) 0           decoder_stage2b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3_concat (Concaten (None, 112, 112, 128 0           decoder_stage3_upsampling[0][0]  \n",
      "                                                                 relu0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_conv (Conv2D)   (None, 112, 112, 32) 36864       decoder_stage3_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_bn (BatchNormal (None, 112, 112, 32) 128         decoder_stage3a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_relu (Activatio (None, 112, 112, 32) 0           decoder_stage3a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_conv (Conv2D)   (None, 112, 112, 32) 9216        decoder_stage3a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_bn (BatchNormal (None, 112, 112, 32) 128         decoder_stage3b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_relu (Activatio (None, 112, 112, 32) 0           decoder_stage3b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4_upsampling (UpSa (None, 224, 224, 32) 0           decoder_stage3b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_conv (Conv2D)   (None, 224, 224, 16) 4608        decoder_stage4_upsampling[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_bn (BatchNormal (None, 224, 224, 16) 64          decoder_stage4a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_relu (Activatio (None, 224, 224, 16) 0           decoder_stage4a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_conv (Conv2D)   (None, 224, 224, 16) 2304        decoder_stage4a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_bn (BatchNormal (None, 224, 224, 16) 64          decoder_stage4b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_relu (Activatio (None, 224, 224, 16) 0           decoder_stage4b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "final_conv (Conv2D)             (None, 224, 224, 2)  290         decoder_stage4b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "softmax (Activation)            (None, 224, 224, 2)  0           final_conv[0][0]                 \n",
      "==================================================================================================\n",
      "Total params: 24,456,299\n",
      "Trainable params: 16,274,405\n",
      "Non-trainable params: 8,181,894\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "BACKBONE = 'resnet34'\n",
    "model = sm.Unet(BACKBONE, encoder_weights='imagenet', classes=2, activation='softmax', encoder_freeze=True, input_shape=(224, 224, 3))\n",
    "for i in range(len(model.layers)):\n",
    "    if model.layers[i].name.startswith('stage4'):\n",
    "        model.layers[i].trainable = True\n",
    "#    if 'f_drop' in layer.name:\n",
    "#        print(layer.name)\n",
    "model.load_weights('resnet34based.hdf5')\n",
    "opt = AdamP(learning_rate=0.01, beta_1=0.9, beta_2=0.999, weight_decay=1e-5)\n",
    "model.compile(\n",
    "    opt,\n",
    "    loss=sm.losses.BinaryFocalLoss(),\n",
    "    metrics=[sm.metrics.IOUScore(threshold=0.5)],\n",
    ")\n",
    "# fit model\n",
    "# if you use data generator use model.fit_generator(...) instead of model.fit(...)\n",
    "# more about `fit_generator` here: https://keras.io/models/sequential/#fit_generator\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0115 - iou_score: 0.8759\n",
      "Epoch 2/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0119 - iou_score: 0.8727\n",
      "Epoch 3/50\n",
      "523/523 [==============================] - 67s 127ms/step - loss: 0.0119 - iou_score: 0.8715\n",
      "Epoch 4/50\n",
      "523/523 [==============================] - 66s 127ms/step - loss: 0.0120 - iou_score: 0.8713\n",
      "Epoch 5/50\n",
      "523/523 [==============================] - 66s 126ms/step - loss: 0.0119 - iou_score: 0.8725\n",
      "Epoch 6/50\n",
      "523/523 [==============================] - 68s 130ms/step - loss: 0.0121 - iou_score: 0.8703\n",
      "Epoch 7/50\n",
      "523/523 [==============================] - 65s 125ms/step - loss: 0.0118 - iou_score: 0.8736\n",
      "Epoch 8/50\n",
      "523/523 [==============================] - 66s 126ms/step - loss: 0.0118 - iou_score: 0.8736\n",
      "Epoch 9/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0118 - iou_score: 0.8728\n",
      "Epoch 10/50\n",
      "523/523 [==============================] - 66s 125ms/step - loss: 0.0119 - iou_score: 0.8722\n",
      "Epoch 11/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0117 - iou_score: 0.8729\n",
      "Epoch 12/50\n",
      "523/523 [==============================] - 66s 126ms/step - loss: 0.0118 - iou_score: 0.8722\n",
      "Epoch 13/50\n",
      "523/523 [==============================] - 64s 122ms/step - loss: 0.0120 - iou_score: 0.8715\n",
      "Epoch 14/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0121 - iou_score: 0.8697\n",
      "Epoch 15/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0118 - iou_score: 0.8719\n",
      "Epoch 16/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0119 - iou_score: 0.8729\n",
      "Epoch 17/50\n",
      "523/523 [==============================] - 63s 121ms/step - loss: 0.0120 - iou_score: 0.8712\n",
      "Epoch 18/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0120 - iou_score: 0.8717\n",
      "Epoch 19/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0120 - iou_score: 0.8719\n",
      "Epoch 20/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0119 - iou_score: 0.8719\n",
      "Epoch 21/50\n",
      "523/523 [==============================] - ETA: 0s - loss: 0.0118 - iou_score: 0.8731\n",
      "Epoch 00021: ReduceLROnPlateau reducing learning rate to 0.004999999888241291.\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0118 - iou_score: 0.8731\n",
      "Epoch 22/50\n",
      "523/523 [==============================] - 64s 122ms/step - loss: 0.0114 - iou_score: 0.8772\n",
      "Epoch 23/50\n",
      "523/523 [==============================] - 65s 125ms/step - loss: 0.0113 - iou_score: 0.8778\n",
      "Epoch 24/50\n",
      "523/523 [==============================] - 66s 126ms/step - loss: 0.0112 - iou_score: 0.8780\n",
      "Epoch 25/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0113 - iou_score: 0.8782\n",
      "Epoch 26/50\n",
      "523/523 [==============================] - 66s 125ms/step - loss: 0.0113 - iou_score: 0.8782\n",
      "Epoch 27/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0113 - iou_score: 0.8791\n",
      "Epoch 28/50\n",
      "523/523 [==============================] - 64s 122ms/step - loss: 0.0112 - iou_score: 0.8787\n",
      "Epoch 29/50\n",
      "523/523 [==============================] - 67s 129ms/step - loss: 0.0112 - iou_score: 0.8785\n",
      "Epoch 30/50\n",
      "523/523 [==============================] - 66s 126ms/step - loss: 0.0110 - iou_score: 0.8806\n",
      "Epoch 31/50\n",
      "523/523 [==============================] - 65s 125ms/step - loss: 0.0113 - iou_score: 0.8782\n",
      "Epoch 32/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0113 - iou_score: 0.8787\n",
      "Epoch 33/50\n",
      "523/523 [==============================] - 63s 120ms/step - loss: 0.0111 - iou_score: 0.8798\n",
      "Epoch 34/50\n",
      "523/523 [==============================] - 65s 123ms/step - loss: 0.0111 - iou_score: 0.8793\n",
      "Epoch 35/50\n",
      "523/523 [==============================] - 63s 121ms/step - loss: 0.0111 - iou_score: 0.8800\n",
      "Epoch 36/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0111 - iou_score: 0.8804\n",
      "Epoch 37/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0111 - iou_score: 0.8795\n",
      "Epoch 38/50\n",
      "523/523 [==============================] - 63s 121ms/step - loss: 0.0112 - iou_score: 0.8794\n",
      "Epoch 39/50\n",
      "523/523 [==============================] - 65s 125ms/step - loss: 0.0111 - iou_score: 0.8799\n",
      "Epoch 40/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0111 - iou_score: 0.8795\n",
      "Epoch 41/50\n",
      "523/523 [==============================] - 63s 121ms/step - loss: 0.0110 - iou_score: 0.8811\n",
      "Epoch 42/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0110 - iou_score: 0.8820\n",
      "Epoch 43/50\n",
      "523/523 [==============================] - 64s 123ms/step - loss: 0.0110 - iou_score: 0.8815\n",
      "Epoch 44/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0110 - iou_score: 0.8811\n",
      "Epoch 45/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0110 - iou_score: 0.8809\n",
      "Epoch 46/50\n",
      "523/523 [==============================] - 65s 125ms/step - loss: 0.0110 - iou_score: 0.8805\n",
      "Epoch 47/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0110 - iou_score: 0.8804\n",
      "Epoch 48/50\n",
      "523/523 [==============================] - 65s 124ms/step - loss: 0.0110 - iou_score: 0.8816\n",
      "Epoch 49/50\n",
      "523/523 [==============================] - 65s 125ms/step - loss: 0.0110 - iou_score: 0.8808\n",
      "Epoch 50/50\n",
      "523/523 [==============================] - ETA: 0s - loss: 0.0110 - iou_score: 0.8810\n",
      "Epoch 00050: ReduceLROnPlateau reducing learning rate to 0.0024999999441206455.\n",
      "523/523 [==============================] - 65s 123ms/step - loss: 0.0110 - iou_score: 0.8810\n"
     ]
    }
   ],
   "source": [
    "batch_size = 16\n",
    "reducelr = keras.callbacks.ReduceLROnPlateau(monitor = 'loss', factor = 0.5, patience = 20, verbose = 1)\n",
    "h = model.fit(generator(X, Y, batch_size),\n",
    "              steps_per_epoch=X.shape[0]//batch_size, epochs=50,\n",
    "              callbacks=[reducelr])# validation_batch_size=64)\n",
    "model.save('resnet34based_finetune.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0229 - iou_score: 0.7518WARNING:tensorflow:Callbacks method `on_test_batch_end` is slow compared to the batch time (batch time: 0.0166s vs `on_test_batch_end` time: 0.0478s). Check your callbacks.\n",
      "471/471 [==============================] - 121s 257ms/step - loss: 0.0229 - iou_score: 0.7518 - val_loss: 0.0496 - val_iou_score: 0.5479\n",
      "Epoch 2/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0170 - iou_score: 0.8209 - val_loss: 0.0526 - val_iou_score: 0.5311\n",
      "Epoch 3/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0165 - iou_score: 0.8275 - val_loss: 0.0522 - val_iou_score: 0.7448\n",
      "Epoch 4/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0157 - iou_score: 0.8337 - val_loss: 0.0139 - val_iou_score: 0.8587\n",
      "Epoch 5/300\n",
      "471/471 [==============================] - 62s 131ms/step - loss: 0.0157 - iou_score: 0.8336 - val_loss: 0.0135 - val_iou_score: 0.8659\n",
      "Epoch 6/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0150 - iou_score: 0.8416 - val_loss: 0.0146 - val_iou_score: 0.8281\n",
      "Epoch 7/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0157 - iou_score: 0.8340 - val_loss: 0.0210 - val_iou_score: 0.7944\n",
      "Epoch 8/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0154 - iou_score: 0.8366 - val_loss: 0.0141 - val_iou_score: 0.8370\n",
      "Epoch 9/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0152 - iou_score: 0.8395 - val_loss: 0.0186 - val_iou_score: 0.8283\n",
      "Epoch 10/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0150 - iou_score: 0.8402 - val_loss: 0.0121 - val_iou_score: 0.8687\n",
      "Epoch 11/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0148 - iou_score: 0.8413 - val_loss: 0.0124 - val_iou_score: 0.8707\n",
      "Epoch 12/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0151 - iou_score: 0.8408 - val_loss: 0.0218 - val_iou_score: 0.8022\n",
      "Epoch 13/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0147 - iou_score: 0.8431 - val_loss: 0.0124 - val_iou_score: 0.8755\n",
      "Epoch 14/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0147 - iou_score: 0.8438 - val_loss: 0.0136 - val_iou_score: 0.8462\n",
      "Epoch 15/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0144 - iou_score: 0.8478 - val_loss: 0.0133 - val_iou_score: 0.8708\n",
      "Epoch 16/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0144 - iou_score: 0.8448 - val_loss: 0.0129 - val_iou_score: 0.8603\n",
      "Epoch 17/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0144 - iou_score: 0.8463 - val_loss: 0.0136 - val_iou_score: 0.8710\n",
      "Epoch 18/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0142 - iou_score: 0.8491 - val_loss: 0.0127 - val_iou_score: 0.8685\n",
      "Epoch 19/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0142 - iou_score: 0.8486 - val_loss: 0.0163 - val_iou_score: 0.8356\n",
      "Epoch 20/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0141 - iou_score: 0.8497 - val_loss: 0.0128 - val_iou_score: 0.8674\n",
      "Epoch 21/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0138 - iou_score: 0.8529 - val_loss: 0.0114 - val_iou_score: 0.8735\n",
      "Epoch 22/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0146 - iou_score: 0.8456 - val_loss: 0.0151 - val_iou_score: 0.8588\n",
      "Epoch 23/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0142 - iou_score: 0.8482 - val_loss: 0.0124 - val_iou_score: 0.8700\n",
      "Epoch 24/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0139 - iou_score: 0.8497 - val_loss: 0.0172 - val_iou_score: 0.8496\n",
      "Epoch 25/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0138 - iou_score: 0.8539 - val_loss: 0.0117 - val_iou_score: 0.8794\n",
      "Epoch 26/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0138 - iou_score: 0.8518 - val_loss: 0.0122 - val_iou_score: 0.8676\n",
      "Epoch 27/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0138 - iou_score: 0.8533 - val_loss: 0.0127 - val_iou_score: 0.8651\n",
      "Epoch 28/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0140 - iou_score: 0.8504 - val_loss: 0.0130 - val_iou_score: 0.8691\n",
      "Epoch 29/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0137 - iou_score: 0.8534 - val_loss: 0.0241 - val_iou_score: 0.7787\n",
      "Epoch 30/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0139 - iou_score: 0.8507 - val_loss: 0.0119 - val_iou_score: 0.8781\n",
      "Epoch 31/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0135 - iou_score: 0.8554 - val_loss: 0.0120 - val_iou_score: 0.8643\n",
      "Epoch 32/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0137 - iou_score: 0.8529 - val_loss: 0.0129 - val_iou_score: 0.8786\n",
      "Epoch 33/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0136 - iou_score: 0.8547 - val_loss: 0.0113 - val_iou_score: 0.8802\n",
      "Epoch 34/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0136 - iou_score: 0.8546 - val_loss: 0.0111 - val_iou_score: 0.8750\n",
      "Epoch 35/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0135 - iou_score: 0.8558 - val_loss: 0.0128 - val_iou_score: 0.8724\n",
      "Epoch 36/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0133 - iou_score: 0.8589 - val_loss: 0.0114 - val_iou_score: 0.8693\n",
      "Epoch 37/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0134 - iou_score: 0.8567 - val_loss: 0.0134 - val_iou_score: 0.8490\n",
      "Epoch 38/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0132 - iou_score: 0.8596 - val_loss: 0.0117 - val_iou_score: 0.8725\n",
      "Epoch 39/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0133 - iou_score: 0.8580 - val_loss: 0.0117 - val_iou_score: 0.8835\n",
      "Epoch 40/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0131 - iou_score: 0.8595 - val_loss: 0.0115 - val_iou_score: 0.8741\n",
      "Epoch 41/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0134 - iou_score: 0.8572 - val_loss: 0.0112 - val_iou_score: 0.8815\n",
      "Epoch 42/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0130 - iou_score: 0.8605 - val_loss: 0.0108 - val_iou_score: 0.8796\n",
      "Epoch 43/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0133 - iou_score: 0.8587 - val_loss: 0.0113 - val_iou_score: 0.8792\n",
      "Epoch 44/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0129 - iou_score: 0.8618 - val_loss: 0.0124 - val_iou_score: 0.8588\n",
      "Epoch 45/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0131 - iou_score: 0.8596 - val_loss: 0.0116 - val_iou_score: 0.8747\n",
      "Epoch 46/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0130 - iou_score: 0.8608 - val_loss: 0.0108 - val_iou_score: 0.8859\n",
      "Epoch 47/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0133 - iou_score: 0.8580 - val_loss: 0.0112 - val_iou_score: 0.8755\n",
      "Epoch 48/300\n",
      "471/471 [==============================] - 61s 129ms/step - loss: 0.0130 - iou_score: 0.8615 - val_loss: 0.0150 - val_iou_score: 0.8599\n",
      "Epoch 49/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0131 - iou_score: 0.8609 - val_loss: 0.0154 - val_iou_score: 0.8473\n",
      "Epoch 50/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0130 - iou_score: 0.8615 - val_loss: 0.0110 - val_iou_score: 0.8791\n",
      "Epoch 51/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0130 - iou_score: 0.8608 - val_loss: 0.0127 - val_iou_score: 0.8769\n",
      "Epoch 52/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0129 - iou_score: 0.8615 - val_loss: 0.0118 - val_iou_score: 0.8763\n",
      "Epoch 53/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0130 - iou_score: 0.8621 - val_loss: 0.0137 - val_iou_score: 0.8578\n",
      "Epoch 54/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0130 - iou_score: 0.8618 - val_loss: 0.0121 - val_iou_score: 0.8775\n",
      "Epoch 55/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0128 - iou_score: 0.8621 - val_loss: 0.0106 - val_iou_score: 0.8831\n",
      "Epoch 56/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0129 - iou_score: 0.8620 - val_loss: 0.0144 - val_iou_score: 0.8551\n",
      "Epoch 57/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0130 - iou_score: 0.8628 - val_loss: 0.0134 - val_iou_score: 0.8665\n",
      "Epoch 58/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0133 - iou_score: 0.8576 - val_loss: 0.0136 - val_iou_score: 0.8737\n",
      "Epoch 59/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0127 - iou_score: 0.8635 - val_loss: 0.0108 - val_iou_score: 0.8853\n",
      "Epoch 60/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0128 - iou_score: 0.8641 - val_loss: 0.0119 - val_iou_score: 0.8717\n",
      "Epoch 61/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0127 - iou_score: 0.8649 - val_loss: 0.0113 - val_iou_score: 0.8818\n",
      "Epoch 62/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0126 - iou_score: 0.8653 - val_loss: 0.0107 - val_iou_score: 0.8844\n",
      "Epoch 63/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0126 - iou_score: 0.8654 - val_loss: 0.0120 - val_iou_score: 0.8747\n",
      "Epoch 64/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0126 - iou_score: 0.8657 - val_loss: 0.0111 - val_iou_score: 0.8822\n",
      "Epoch 65/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0128 - iou_score: 0.8647 - val_loss: 0.0114 - val_iou_score: 0.8738\n",
      "Epoch 66/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0125 - iou_score: 0.8660 - val_loss: 0.0105 - val_iou_score: 0.8872\n",
      "Epoch 67/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0124 - iou_score: 0.8672 - val_loss: 0.0109 - val_iou_score: 0.8799\n",
      "Epoch 68/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0125 - iou_score: 0.8669 - val_loss: 0.0129 - val_iou_score: 0.8823\n",
      "Epoch 69/300\n",
      "471/471 [==============================] - 58s 122ms/step - loss: 0.0127 - iou_score: 0.8659 - val_loss: 0.0104 - val_iou_score: 0.8867\n",
      "Epoch 70/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0123 - iou_score: 0.8681 - val_loss: 0.0102 - val_iou_score: 0.8889\n",
      "Epoch 71/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0124 - iou_score: 0.8661 - val_loss: 0.0112 - val_iou_score: 0.8812\n",
      "Epoch 72/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0126 - iou_score: 0.8661 - val_loss: 0.0222 - val_iou_score: 0.7940\n",
      "Epoch 73/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0125 - iou_score: 0.8657 - val_loss: 0.0106 - val_iou_score: 0.8836\n",
      "Epoch 74/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0125 - iou_score: 0.8664 - val_loss: 0.0157 - val_iou_score: 0.8412\n",
      "Epoch 75/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0123 - iou_score: 0.8678 - val_loss: 0.0109 - val_iou_score: 0.8861\n",
      "Epoch 76/300\n",
      "471/471 [==============================] - 61s 129ms/step - loss: 0.0122 - iou_score: 0.8693 - val_loss: 0.0107 - val_iou_score: 0.8850\n",
      "Epoch 77/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0122 - iou_score: 0.8690 - val_loss: 0.0103 - val_iou_score: 0.8892\n",
      "Epoch 78/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0123 - iou_score: 0.8683 - val_loss: 0.0104 - val_iou_score: 0.8859\n",
      "Epoch 79/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0121 - iou_score: 0.8707 - val_loss: 0.0106 - val_iou_score: 0.8814\n",
      "Epoch 80/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0124 - iou_score: 0.8677 - val_loss: 0.0132 - val_iou_score: 0.8824\n",
      "Epoch 81/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0124 - iou_score: 0.8669 - val_loss: 0.0113 - val_iou_score: 0.8805\n",
      "Epoch 82/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0120 - iou_score: 0.8709 - val_loss: 0.0103 - val_iou_score: 0.8886\n",
      "Epoch 83/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0123 - iou_score: 0.8679 - val_loss: 0.0111 - val_iou_score: 0.8831\n",
      "Epoch 84/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0124 - iou_score: 0.8679 - val_loss: 0.0111 - val_iou_score: 0.8848\n",
      "Epoch 85/300\n",
      "471/471 [==============================] - 57s 122ms/step - loss: 0.0121 - iou_score: 0.8703 - val_loss: 0.0121 - val_iou_score: 0.8619\n",
      "Epoch 86/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0123 - iou_score: 0.8671 - val_loss: 0.0110 - val_iou_score: 0.8809\n",
      "Epoch 87/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0124 - iou_score: 0.8673 - val_loss: 0.0114 - val_iou_score: 0.8713\n",
      "Epoch 88/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0121 - iou_score: 0.8699 - val_loss: 0.0101 - val_iou_score: 0.8901\n",
      "Epoch 89/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0126 - iou_score: 0.8669 - val_loss: 0.0114 - val_iou_score: 0.8853\n",
      "Epoch 90/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0124 - iou_score: 0.8673\n",
      "Epoch 00090: ReduceLROnPlateau reducing learning rate to 0.004999999888241291.\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0124 - iou_score: 0.8673 - val_loss: 0.0121 - val_iou_score: 0.8797\n",
      "Epoch 91/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0118 - iou_score: 0.8724 - val_loss: 0.0098 - val_iou_score: 0.8925\n",
      "Epoch 92/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0118 - iou_score: 0.8735 - val_loss: 0.0101 - val_iou_score: 0.8908\n",
      "Epoch 93/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0117 - iou_score: 0.8751 - val_loss: 0.0098 - val_iou_score: 0.8906\n",
      "Epoch 94/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0115 - iou_score: 0.8766 - val_loss: 0.0099 - val_iou_score: 0.8934\n",
      "Epoch 95/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0114 - iou_score: 0.8770 - val_loss: 0.0099 - val_iou_score: 0.8949\n",
      "Epoch 96/300\n",
      "471/471 [==============================] - 61s 129ms/step - loss: 0.0115 - iou_score: 0.8761 - val_loss: 0.0106 - val_iou_score: 0.8804\n",
      "Epoch 97/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0115 - iou_score: 0.8766 - val_loss: 0.0102 - val_iou_score: 0.8868\n",
      "Epoch 98/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0114 - iou_score: 0.8761 - val_loss: 0.0099 - val_iou_score: 0.8955\n",
      "Epoch 99/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0114 - iou_score: 0.8771 - val_loss: 0.0097 - val_iou_score: 0.8959\n",
      "Epoch 100/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0114 - iou_score: 0.8768 - val_loss: 0.0099 - val_iou_score: 0.8971\n",
      "Epoch 101/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0112 - iou_score: 0.8789 - val_loss: 0.0097 - val_iou_score: 0.8922\n",
      "Epoch 102/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0114 - iou_score: 0.8773 - val_loss: 0.0099 - val_iou_score: 0.8907\n",
      "Epoch 103/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0112 - iou_score: 0.8771 - val_loss: 0.0098 - val_iou_score: 0.8904\n",
      "Epoch 104/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0115 - iou_score: 0.8766 - val_loss: 0.0097 - val_iou_score: 0.8948\n",
      "Epoch 105/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0114 - iou_score: 0.8778 - val_loss: 0.0101 - val_iou_score: 0.8939\n",
      "Epoch 106/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0114 - iou_score: 0.8778 - val_loss: 0.0094 - val_iou_score: 0.8977\n",
      "Epoch 107/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0113 - iou_score: 0.8786 - val_loss: 0.0098 - val_iou_score: 0.8952\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 108/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0113 - iou_score: 0.8777 - val_loss: 0.0096 - val_iou_score: 0.8938\n",
      "Epoch 109/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0113 - iou_score: 0.8793 - val_loss: 0.0098 - val_iou_score: 0.8905\n",
      "Epoch 110/300\n",
      "471/471 [==============================] - 61s 129ms/step - loss: 0.0113 - iou_score: 0.8789 - val_loss: 0.0097 - val_iou_score: 0.8947\n",
      "Epoch 111/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0112 - iou_score: 0.8782 - val_loss: 0.0095 - val_iou_score: 0.8953\n",
      "Epoch 112/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0112 - iou_score: 0.8794 - val_loss: 0.0097 - val_iou_score: 0.8910\n",
      "Epoch 113/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0113 - iou_score: 0.8786 - val_loss: 0.0095 - val_iou_score: 0.8957\n",
      "Epoch 114/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0111 - iou_score: 0.8799 - val_loss: 0.0099 - val_iou_score: 0.8889\n",
      "Epoch 115/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0112 - iou_score: 0.8789 - val_loss: 0.0096 - val_iou_score: 0.8965\n",
      "Epoch 116/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0112 - iou_score: 0.8795 - val_loss: 0.0096 - val_iou_score: 0.8966\n",
      "Epoch 117/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0113 - iou_score: 0.8784 - val_loss: 0.0097 - val_iou_score: 0.8946\n",
      "Epoch 118/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0110 - iou_score: 0.8800 - val_loss: 0.0099 - val_iou_score: 0.8957\n",
      "Epoch 119/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0113 - iou_score: 0.8791 - val_loss: 0.0097 - val_iou_score: 0.8922\n",
      "Epoch 120/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0113 - iou_score: 0.8785 - val_loss: 0.0101 - val_iou_score: 0.8908\n",
      "Epoch 121/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0118 - iou_score: 0.8727 - val_loss: 0.0096 - val_iou_score: 0.8941\n",
      "Epoch 122/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0117 - iou_score: 0.8753 - val_loss: 0.0097 - val_iou_score: 0.8927\n",
      "Epoch 123/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0114 - iou_score: 0.8765 - val_loss: 0.0096 - val_iou_score: 0.8965\n",
      "Epoch 124/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0113 - iou_score: 0.8777 - val_loss: 0.0098 - val_iou_score: 0.8915\n",
      "Epoch 125/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0112 - iou_score: 0.8791 - val_loss: 0.0097 - val_iou_score: 0.8966\n",
      "Epoch 126/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0113 - iou_score: 0.8789\n",
      "Epoch 00126: ReduceLROnPlateau reducing learning rate to 0.0024999999441206455.\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0113 - iou_score: 0.8789 - val_loss: 0.0095 - val_iou_score: 0.8975\n",
      "Epoch 127/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0110 - iou_score: 0.8816 - val_loss: 0.0093 - val_iou_score: 0.8978\n",
      "Epoch 128/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0110 - iou_score: 0.8806 - val_loss: 0.0099 - val_iou_score: 0.8905\n",
      "Epoch 129/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0109 - iou_score: 0.8825 - val_loss: 0.0095 - val_iou_score: 0.8961\n",
      "Epoch 130/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0110 - iou_score: 0.8811 - val_loss: 0.0094 - val_iou_score: 0.8958\n",
      "Epoch 131/300\n",
      "471/471 [==============================] - 58s 122ms/step - loss: 0.0109 - iou_score: 0.8825 - val_loss: 0.0095 - val_iou_score: 0.8964\n",
      "Epoch 132/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0111 - iou_score: 0.8810 - val_loss: 0.0093 - val_iou_score: 0.8959\n",
      "Epoch 133/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0109 - iou_score: 0.8810 - val_loss: 0.0093 - val_iou_score: 0.8997\n",
      "Epoch 134/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0110 - iou_score: 0.8810 - val_loss: 0.0094 - val_iou_score: 0.8947\n",
      "Epoch 135/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0108 - iou_score: 0.8829 - val_loss: 0.0092 - val_iou_score: 0.8999\n",
      "Epoch 136/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0107 - iou_score: 0.8834 - val_loss: 0.0093 - val_iou_score: 0.8999\n",
      "Epoch 137/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0109 - iou_score: 0.8819 - val_loss: 0.0092 - val_iou_score: 0.9008\n",
      "Epoch 138/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0109 - iou_score: 0.8828 - val_loss: 0.0093 - val_iou_score: 0.9006\n",
      "Epoch 139/300\n",
      "471/471 [==============================] - 57s 122ms/step - loss: 0.0109 - iou_score: 0.8835 - val_loss: 0.0094 - val_iou_score: 0.8972\n",
      "Epoch 140/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0107 - iou_score: 0.8838 - val_loss: 0.0094 - val_iou_score: 0.8999\n",
      "Epoch 141/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0109 - iou_score: 0.8820 - val_loss: 0.0093 - val_iou_score: 0.9005\n",
      "Epoch 142/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0107 - iou_score: 0.8842 - val_loss: 0.0094 - val_iou_score: 0.8988\n",
      "Epoch 143/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0109 - iou_score: 0.8819 - val_loss: 0.0092 - val_iou_score: 0.8999\n",
      "Epoch 144/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0108 - iou_score: 0.8830 - val_loss: 0.0091 - val_iou_score: 0.9004\n",
      "Epoch 145/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0108 - iou_score: 0.8836 - val_loss: 0.0096 - val_iou_score: 0.8930\n",
      "Epoch 146/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0108 - iou_score: 0.8826 - val_loss: 0.0092 - val_iou_score: 0.9007\n",
      "Epoch 147/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0108 - iou_score: 0.8838 - val_loss: 0.0092 - val_iou_score: 0.8999\n",
      "Epoch 148/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0109 - iou_score: 0.8838 - val_loss: 0.0096 - val_iou_score: 0.8971\n",
      "Epoch 149/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0107 - iou_score: 0.8829 - val_loss: 0.0092 - val_iou_score: 0.8987\n",
      "Epoch 150/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0108 - iou_score: 0.8830 - val_loss: 0.0101 - val_iou_score: 0.8951\n",
      "Epoch 151/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0108 - iou_score: 0.8845 - val_loss: 0.0093 - val_iou_score: 0.8997\n",
      "Epoch 152/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0108 - iou_score: 0.8833 - val_loss: 0.0091 - val_iou_score: 0.8999\n",
      "Epoch 153/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0107 - iou_score: 0.8842 - val_loss: 0.0092 - val_iou_score: 0.9006\n",
      "Epoch 154/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0107 - iou_score: 0.8844 - val_loss: 0.0091 - val_iou_score: 0.9001\n",
      "Epoch 155/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0107 - iou_score: 0.8842\n",
      "Epoch 00155: ReduceLROnPlateau reducing learning rate to 0.0012499999720603228.\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0107 - iou_score: 0.8842 - val_loss: 0.0091 - val_iou_score: 0.9014\n",
      "Epoch 156/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0106 - iou_score: 0.8847 - val_loss: 0.0091 - val_iou_score: 0.9016\n",
      "Epoch 157/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0105 - iou_score: 0.8859 - val_loss: 0.0090 - val_iou_score: 0.9018\n",
      "Epoch 158/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0105 - iou_score: 0.8861 - val_loss: 0.0090 - val_iou_score: 0.9020\n",
      "Epoch 159/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0104 - iou_score: 0.8873 - val_loss: 0.0091 - val_iou_score: 0.9020\n",
      "Epoch 160/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0106 - iou_score: 0.8855 - val_loss: 0.0091 - val_iou_score: 0.9022\n",
      "Epoch 161/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0104 - iou_score: 0.8870 - val_loss: 0.0092 - val_iou_score: 0.8972\n",
      "Epoch 162/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0106 - iou_score: 0.8852 - val_loss: 0.0092 - val_iou_score: 0.8989\n",
      "Epoch 163/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0105 - iou_score: 0.8870 - val_loss: 0.0091 - val_iou_score: 0.9001\n",
      "Epoch 164/300\n",
      "471/471 [==============================] - 58s 122ms/step - loss: 0.0105 - iou_score: 0.8857 - val_loss: 0.0091 - val_iou_score: 0.9017\n",
      "Epoch 165/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0106 - iou_score: 0.8848 - val_loss: 0.0090 - val_iou_score: 0.9006\n",
      "Epoch 166/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0104 - iou_score: 0.8869 - val_loss: 0.0091 - val_iou_score: 0.9020\n",
      "Epoch 167/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0105 - iou_score: 0.8863 - val_loss: 0.0090 - val_iou_score: 0.9019\n",
      "Epoch 168/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0104 - iou_score: 0.8880 - val_loss: 0.0090 - val_iou_score: 0.9017\n",
      "Epoch 169/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8885 - val_loss: 0.0090 - val_iou_score: 0.9005\n",
      "Epoch 170/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0105 - iou_score: 0.8862 - val_loss: 0.0090 - val_iou_score: 0.9025\n",
      "Epoch 171/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0106 - iou_score: 0.8864 - val_loss: 0.0090 - val_iou_score: 0.9005\n",
      "Epoch 172/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0103 - iou_score: 0.8881 - val_loss: 0.0091 - val_iou_score: 0.9018\n",
      "Epoch 173/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0105 - iou_score: 0.8868 - val_loss: 0.0090 - val_iou_score: 0.9005\n",
      "Epoch 174/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0105 - iou_score: 0.8853 - val_loss: 0.0090 - val_iou_score: 0.9017\n",
      "Epoch 175/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0105 - iou_score: 0.8867 - val_loss: 0.0090 - val_iou_score: 0.9018\n",
      "Epoch 176/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0105 - iou_score: 0.8868 - val_loss: 0.0090 - val_iou_score: 0.9018\n",
      "Epoch 177/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0104 - iou_score: 0.8872\n",
      "Epoch 00177: ReduceLROnPlateau reducing learning rate to 0.0006249999860301614.\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0104 - iou_score: 0.8872 - val_loss: 0.0089 - val_iou_score: 0.9030\n",
      "Epoch 178/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0105 - iou_score: 0.8861 - val_loss: 0.0089 - val_iou_score: 0.9029\n",
      "Epoch 179/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0104 - iou_score: 0.8862 - val_loss: 0.0089 - val_iou_score: 0.9023\n",
      "Epoch 180/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8881 - val_loss: 0.0090 - val_iou_score: 0.9027\n",
      "Epoch 181/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0103 - iou_score: 0.8880 - val_loss: 0.0089 - val_iou_score: 0.9031\n",
      "Epoch 182/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0104 - iou_score: 0.8889 - val_loss: 0.0090 - val_iou_score: 0.9032\n",
      "Epoch 183/300\n",
      "471/471 [==============================] - 61s 129ms/step - loss: 0.0103 - iou_score: 0.8875 - val_loss: 0.0089 - val_iou_score: 0.9027\n",
      "Epoch 184/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0105 - iou_score: 0.8877 - val_loss: 0.0089 - val_iou_score: 0.9027\n",
      "Epoch 185/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0102 - iou_score: 0.8891 - val_loss: 0.0090 - val_iou_score: 0.9028\n",
      "Epoch 186/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0104 - iou_score: 0.8866 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 187/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0103 - iou_score: 0.8887 - val_loss: 0.0090 - val_iou_score: 0.9019\n",
      "Epoch 188/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0103 - iou_score: 0.8881 - val_loss: 0.0089 - val_iou_score: 0.9029\n",
      "Epoch 189/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0103 - iou_score: 0.8887 - val_loss: 0.0089 - val_iou_score: 0.9029\n",
      "Epoch 190/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0103 - iou_score: 0.8884 - val_loss: 0.0090 - val_iou_score: 0.9027\n",
      "Epoch 191/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0104 - iou_score: 0.8878 - val_loss: 0.0089 - val_iou_score: 0.9022\n",
      "Epoch 192/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0104 - iou_score: 0.8875 - val_loss: 0.0089 - val_iou_score: 0.9031\n",
      "Epoch 193/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0103 - iou_score: 0.8874 - val_loss: 0.0089 - val_iou_score: 0.9031\n",
      "Epoch 194/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8886 - val_loss: 0.0089 - val_iou_score: 0.9039\n",
      "Epoch 195/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8887 - val_loss: 0.0089 - val_iou_score: 0.9031\n",
      "Epoch 196/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0102 - iou_score: 0.8887 - val_loss: 0.0090 - val_iou_score: 0.9013\n",
      "Epoch 197/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0104 - iou_score: 0.8872 - val_loss: 0.0089 - val_iou_score: 0.9025\n",
      "Epoch 198/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0101 - iou_score: 0.8893\n",
      "Epoch 00198: ReduceLROnPlateau reducing learning rate to 0.0003124999930150807.\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0101 - iou_score: 0.8893 - val_loss: 0.0089 - val_iou_score: 0.9028\n",
      "Epoch 199/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0104 - iou_score: 0.8882 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 200/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0103 - iou_score: 0.8884 - val_loss: 0.0089 - val_iou_score: 0.9030\n",
      "Epoch 201/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8882 - val_loss: 0.0089 - val_iou_score: 0.9033\n",
      "Epoch 202/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8885 - val_loss: 0.0089 - val_iou_score: 0.9030\n",
      "Epoch 203/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8874 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 204/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8878 - val_loss: 0.0089 - val_iou_score: 0.9034\n",
      "Epoch 205/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0103 - iou_score: 0.8889 - val_loss: 0.0089 - val_iou_score: 0.9034\n",
      "Epoch 206/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0102 - iou_score: 0.8889 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 207/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8900 - val_loss: 0.0088 - val_iou_score: 0.9040\n",
      "Epoch 208/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8887 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 209/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8900 - val_loss: 0.0089 - val_iou_score: 0.9034\n",
      "Epoch 210/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8893 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 211/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0103 - iou_score: 0.8890 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 212/300\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8899 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 213/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8896 - val_loss: 0.0089 - val_iou_score: 0.9033\n",
      "Epoch 214/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0102 - iou_score: 0.8889 - val_loss: 0.0089 - val_iou_score: 0.9034\n",
      "Epoch 215/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0103 - iou_score: 0.8891 - val_loss: 0.0089 - val_iou_score: 0.9030\n",
      "Epoch 216/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8895 - val_loss: 0.0089 - val_iou_score: 0.9032\n",
      "Epoch 217/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0102 - iou_score: 0.8887 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 218/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0103 - iou_score: 0.8891\n",
      "Epoch 00218: ReduceLROnPlateau reducing learning rate to 0.00015624999650754035.\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0103 - iou_score: 0.8891 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 219/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8897 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 220/300\n",
      "471/471 [==============================] - 61s 129ms/step - loss: 0.0102 - iou_score: 0.8893 - val_loss: 0.0089 - val_iou_score: 0.9032\n",
      "Epoch 221/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0103 - iou_score: 0.8878 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 222/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8897 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 223/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8891 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 224/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0103 - iou_score: 0.8892 - val_loss: 0.0089 - val_iou_score: 0.9033\n",
      "Epoch 225/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8883 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 226/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8881 - val_loss: 0.0088 - val_iou_score: 0.9036\n",
      "Epoch 227/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8897 - val_loss: 0.0088 - val_iou_score: 0.9041\n",
      "Epoch 228/300\n",
      "471/471 [==============================] - 58s 122ms/step - loss: 0.0102 - iou_score: 0.8889 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 229/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0102 - iou_score: 0.8886 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 230/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8893 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 231/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8891 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 232/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0103 - iou_score: 0.8890 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 233/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8905 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 234/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8886 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 235/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0101 - iou_score: 0.8897 - val_loss: 0.0089 - val_iou_score: 0.9034\n",
      "Epoch 236/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0103 - iou_score: 0.8886 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 237/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8893 - val_loss: 0.0089 - val_iou_score: 0.9039\n",
      "Epoch 238/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0103 - iou_score: 0.8894\n",
      "Epoch 00238: ReduceLROnPlateau reducing learning rate to 7.812499825377017e-05.\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0103 - iou_score: 0.8894 - val_loss: 0.0089 - val_iou_score: 0.9031\n",
      "Epoch 239/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0103 - iou_score: 0.8883 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 240/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0101 - iou_score: 0.8899 - val_loss: 0.0089 - val_iou_score: 0.9039\n",
      "Epoch 241/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0102 - iou_score: 0.8898 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 242/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8887 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 243/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8899 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 244/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0102 - iou_score: 0.8891 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 245/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0103 - iou_score: 0.8900 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 246/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0103 - iou_score: 0.8888 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 247/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0100 - iou_score: 0.8902 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 248/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8910 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 249/300\n",
      "471/471 [==============================] - 57s 121ms/step - loss: 0.0102 - iou_score: 0.8905 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 250/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0102 - iou_score: 0.8891 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 251/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8901 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 252/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0102 - iou_score: 0.8896 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 253/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8901 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 254/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0103 - iou_score: 0.8889 - val_loss: 0.0088 - val_iou_score: 0.9036\n",
      "Epoch 255/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0102 - iou_score: 0.8895 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 256/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0100 - iou_score: 0.8907 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 257/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0104 - iou_score: 0.8883 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 258/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0102 - iou_score: 0.8902\n",
      "Epoch 00258: ReduceLROnPlateau reducing learning rate to 3.9062499126885086e-05.\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8902 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 259/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8892 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 260/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8889 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 261/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8900 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 262/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0103 - iou_score: 0.8893 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 263/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0102 - iou_score: 0.8897 - val_loss: 0.0089 - val_iou_score: 0.9035\n",
      "Epoch 264/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0101 - iou_score: 0.8901 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 265/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0102 - iou_score: 0.8903 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 266/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8883 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 267/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0102 - iou_score: 0.8888 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 268/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0103 - iou_score: 0.8893 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 269/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0101 - iou_score: 0.8896 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 270/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8895 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 271/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0101 - iou_score: 0.8900 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 272/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0102 - iou_score: 0.8892 - val_loss: 0.0089 - val_iou_score: 0.9038\n",
      "Epoch 273/300\n",
      "471/471 [==============================] - 60s 128ms/step - loss: 0.0101 - iou_score: 0.8898 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 274/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8892 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 275/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0103 - iou_score: 0.8897 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 276/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8895 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 277/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8905 - val_loss: 0.0089 - val_iou_score: 0.9036\n",
      "Epoch 278/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0102 - iou_score: 0.8894\n",
      "Epoch 00278: ReduceLROnPlateau reducing learning rate to 1.9531249563442543e-05.\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8894 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 279/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8906 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 280/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0101 - iou_score: 0.8894 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 281/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8903 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 282/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8898 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 283/300\n",
      "471/471 [==============================] - 60s 126ms/step - loss: 0.0102 - iou_score: 0.8883 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 284/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8894 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 285/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0102 - iou_score: 0.8886 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 286/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0100 - iou_score: 0.8908 - val_loss: 0.0088 - val_iou_score: 0.9039\n",
      "Epoch 287/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0102 - iou_score: 0.8898 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 288/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0101 - iou_score: 0.8906 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 289/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0101 - iou_score: 0.8898 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 290/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0100 - iou_score: 0.8902 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 291/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0102 - iou_score: 0.8893 - val_loss: 0.0088 - val_iou_score: 0.9039\n",
      "Epoch 292/300\n",
      "471/471 [==============================] - 59s 124ms/step - loss: 0.0102 - iou_score: 0.8902 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 293/300\n",
      "471/471 [==============================] - 59s 126ms/step - loss: 0.0101 - iou_score: 0.8896 - val_loss: 0.0088 - val_iou_score: 0.9039\n",
      "Epoch 294/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0101 - iou_score: 0.8912 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 295/300\n",
      "471/471 [==============================] - 58s 124ms/step - loss: 0.0102 - iou_score: 0.8899 - val_loss: 0.0088 - val_iou_score: 0.9037\n",
      "Epoch 296/300\n",
      "471/471 [==============================] - 58s 123ms/step - loss: 0.0101 - iou_score: 0.8902 - val_loss: 0.0089 - val_iou_score: 0.9037\n",
      "Epoch 297/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8902 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 298/300\n",
      "471/471 [==============================] - ETA: 0s - loss: 0.0101 - iou_score: 0.8903\n",
      "Epoch 00298: ReduceLROnPlateau reducing learning rate to 9.765624781721272e-06.\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0101 - iou_score: 0.8903 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 299/300\n",
      "471/471 [==============================] - 59s 125ms/step - loss: 0.0102 - iou_score: 0.8893 - val_loss: 0.0088 - val_iou_score: 0.9038\n",
      "Epoch 300/300\n",
      "471/471 [==============================] - 60s 127ms/step - loss: 0.0101 - iou_score: 0.8905 - val_loss: 0.0089 - val_iou_score: 0.9037\n"
     ]
    }
   ],
   "source": [
    "batch_size = 16\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.1)\n",
    "reducelr = keras.callbacks.ReduceLROnPlateau(monitor = 'val_loss', factor = 0.5, patience = 20, verbose = 1)\n",
    "h = model.fit(generator(X_train, Y_train, batch_size),\n",
    "              validation_data=(X_test, Y_test),\n",
    "              steps_per_epoch=X_train.shape[0]//batch_size, epochs=300,\n",
    "              callbacks=[reducelr])# validation_batch_size=64)\n",
    "model.save('resnet34based.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('try5.hdf5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Eval"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"functional_3\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "data (InputLayer)               [(None, 224, 224, 3) 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "bn_data (BatchNormalization)    (None, 224, 224, 3)  9           data[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d (ZeroPadding2D)  (None, 230, 230, 3)  0           bn_data[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv0 (Conv2D)                  (None, 112, 112, 64) 9408        zero_padding2d[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "bn0 (BatchNormalization)        (None, 112, 112, 64) 256         conv0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "relu0 (Activation)              (None, 112, 112, 64) 0           bn0[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_1 (ZeroPadding2D (None, 114, 114, 64) 0           relu0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "pooling0 (MaxPooling2D)         (None, 56, 56, 64)   0           zero_padding2d_1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_bn1 (BatchNormaliz (None, 56, 56, 64)   256         pooling0[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_2 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_3 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_3[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_sc (Conv2D)        (None, 56, 56, 64)   4096        stage1_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add (Add)                       (None, 56, 56, 64)   0           stage1_unit1_conv2[0][0]         \n",
      "                                                                 stage1_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_4 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_4[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_5 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_5[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "add_1 (Add)                     (None, 56, 56, 64)   0           stage1_unit2_conv2[0][0]         \n",
      "                                                                 add[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_6 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_6[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_7 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_7[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "add_2 (Add)                     (None, 56, 56, 64)   0           stage1_unit3_conv2[0][0]         \n",
      "                                                                 add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_relu1 (Activation) (None, 56, 56, 64)   0           stage2_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_8 (ZeroPadding2D (None, 58, 58, 64)   0           stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_conv1 (Conv2D)     (None, 28, 28, 128)  73728       zero_padding2d_8[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_9 (ZeroPadding2D (None, 30, 30, 128)  0           stage2_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_9[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_sc (Conv2D)        (None, 28, 28, 128)  8192        stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_3 (Add)                     (None, 28, 28, 128)  0           stage2_unit1_conv2[0][0]         \n",
      "                                                                 stage2_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_10 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_10[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_11 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_11[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_4 (Add)                     (None, 28, 28, 128)  0           stage2_unit2_conv2[0][0]         \n",
      "                                                                 add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_12 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_12[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_13 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_13[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_5 (Add)                     (None, 28, 28, 128)  0           stage2_unit3_conv2[0][0]         \n",
      "                                                                 add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit4_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_14 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit4_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_14[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit4_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit4_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_15 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit4_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_15[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_6 (Add)                     (None, 28, 28, 128)  0           stage2_unit4_conv2[0][0]         \n",
      "                                                                 add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_6[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_relu1 (Activation) (None, 28, 28, 128)  0           stage3_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_16 (ZeroPadding2 (None, 30, 30, 128)  0           stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_conv1 (Conv2D)     (None, 14, 14, 256)  294912      zero_padding2d_16[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_17 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_17[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_sc (Conv2D)        (None, 14, 14, 256)  32768       stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_7 (Add)                     (None, 14, 14, 256)  0           stage3_unit1_conv2[0][0]         \n",
      "                                                                 stage3_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_18 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_18[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_19 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_19[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_8 (Add)                     (None, 14, 14, 256)  0           stage3_unit2_conv2[0][0]         \n",
      "                                                                 add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_20 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_20[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_21 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_21[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_9 (Add)                     (None, 14, 14, 256)  0           stage3_unit3_conv2[0][0]         \n",
      "                                                                 add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit4_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_22 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit4_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_22[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit4_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit4_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_23 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit4_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_23[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_10 (Add)                    (None, 14, 14, 256)  0           stage3_unit4_conv2[0][0]         \n",
      "                                                                 add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit5_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_24 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit5_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_24[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit5_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit5_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_25 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit5_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_25[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_11 (Add)                    (None, 14, 14, 256)  0           stage3_unit5_conv2[0][0]         \n",
      "                                                                 add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit6_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_26 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit6_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_26[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit6_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit6_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_27 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit6_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_27[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_12 (Add)                    (None, 14, 14, 256)  0           stage3_unit6_conv2[0][0]         \n",
      "                                                                 add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_12[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_relu1 (Activation) (None, 14, 14, 256)  0           stage4_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_28 (ZeroPadding2 (None, 16, 16, 256)  0           stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_conv1 (Conv2D)     (None, 7, 7, 512)    1179648     zero_padding2d_28[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_29 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_29[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_sc (Conv2D)        (None, 7, 7, 512)    131072      stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_13 (Add)                    (None, 7, 7, 512)    0           stage4_unit1_conv2[0][0]         \n",
      "                                                                 stage4_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_bn1 (BatchNormaliz (None, 7, 7, 512)    2048        add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_relu1 (Activation) (None, 7, 7, 512)    0           stage4_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_30 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_conv1 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_30[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_31 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_31[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_14 (Add)                    (None, 7, 7, 512)    0           stage4_unit2_conv2[0][0]         \n",
      "                                                                 add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_bn1 (BatchNormaliz (None, 7, 7, 512)    2048        add_14[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_relu1 (Activation) (None, 7, 7, 512)    0           stage4_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_32 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_conv1 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_32[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_33 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_33[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_15 (Add)                    (None, 7, 7, 512)    0           stage4_unit3_conv2[0][0]         \n",
      "                                                                 add_14[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "bn1 (BatchNormalization)        (None, 7, 7, 512)    2048        add_15[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "relu1 (Activation)              (None, 7, 7, 512)    0           bn1[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0_upsampling (UpSa (None, 14, 14, 512)  0           relu1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0_concat (Concaten (None, 14, 14, 768)  0           decoder_stage0_upsampling[0][0]  \n",
      "                                                                 stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_conv (Conv2D)   (None, 14, 14, 256)  1769472     decoder_stage0_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_bn (BatchNormal (None, 14, 14, 256)  1024        decoder_stage0a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_relu (Activatio (None, 14, 14, 256)  0           decoder_stage0a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_conv (Conv2D)   (None, 14, 14, 256)  589824      decoder_stage0a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_bn (BatchNormal (None, 14, 14, 256)  1024        decoder_stage0b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_relu (Activatio (None, 14, 14, 256)  0           decoder_stage0b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1_upsampling (UpSa (None, 28, 28, 256)  0           decoder_stage0b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1_concat (Concaten (None, 28, 28, 384)  0           decoder_stage1_upsampling[0][0]  \n",
      "                                                                 stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_conv (Conv2D)   (None, 28, 28, 128)  442368      decoder_stage1_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_bn (BatchNormal (None, 28, 28, 128)  512         decoder_stage1a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_relu (Activatio (None, 28, 28, 128)  0           decoder_stage1a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_conv (Conv2D)   (None, 28, 28, 128)  147456      decoder_stage1a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_bn (BatchNormal (None, 28, 28, 128)  512         decoder_stage1b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_relu (Activatio (None, 28, 28, 128)  0           decoder_stage1b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2_upsampling (UpSa (None, 56, 56, 128)  0           decoder_stage1b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2_concat (Concaten (None, 56, 56, 192)  0           decoder_stage2_upsampling[0][0]  \n",
      "                                                                 stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_conv (Conv2D)   (None, 56, 56, 64)   110592      decoder_stage2_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_bn (BatchNormal (None, 56, 56, 64)   256         decoder_stage2a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_relu (Activatio (None, 56, 56, 64)   0           decoder_stage2a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_conv (Conv2D)   (None, 56, 56, 64)   36864       decoder_stage2a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_bn (BatchNormal (None, 56, 56, 64)   256         decoder_stage2b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_relu (Activatio (None, 56, 56, 64)   0           decoder_stage2b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3_upsampling (UpSa (None, 112, 112, 64) 0           decoder_stage2b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3_concat (Concaten (None, 112, 112, 128 0           decoder_stage3_upsampling[0][0]  \n",
      "                                                                 relu0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_conv (Conv2D)   (None, 112, 112, 32) 36864       decoder_stage3_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_bn (BatchNormal (None, 112, 112, 32) 128         decoder_stage3a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_relu (Activatio (None, 112, 112, 32) 0           decoder_stage3a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_conv (Conv2D)   (None, 112, 112, 32) 9216        decoder_stage3a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_bn (BatchNormal (None, 112, 112, 32) 128         decoder_stage3b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_relu (Activatio (None, 112, 112, 32) 0           decoder_stage3b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4_upsampling (UpSa (None, 224, 224, 32) 0           decoder_stage3b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_conv (Conv2D)   (None, 224, 224, 16) 4608        decoder_stage4_upsampling[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_bn (BatchNormal (None, 224, 224, 16) 64          decoder_stage4a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_relu (Activatio (None, 224, 224, 16) 0           decoder_stage4a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_conv (Conv2D)   (None, 224, 224, 16) 2304        decoder_stage4a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_bn (BatchNormal (None, 224, 224, 16) 64          decoder_stage4b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_relu (Activatio (None, 224, 224, 16) 0           decoder_stage4b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "final_conv (Conv2D)             (None, 224, 224, 2)  290         decoder_stage4b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "softmax (Activation)            (None, 224, 224, 2)  0           final_conv[0][0]                 \n",
      "==================================================================================================\n",
      "Total params: 24,456,299\n",
      "Trainable params: 3,167,205\n",
      "Non-trainable params: 21,289,094\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "BACKBONE = 'resnet34'\n",
    "model = sm.Unet(BACKBONE, encoder_weights='imagenet', classes=2, activation='softmax', encoder_freeze=True, input_shape=(224, 224, 3))\n",
    "model.summary()\n",
    "model.load_weights('resnet34based_finetune.hdf5')\n",
    "model.compile('adam', 'categorical_crossentropy', metrics='accuracy')\n",
    "model.save('resnet_norm_finetune.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"functional_3\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "data (InputLayer)               [(None, 224, 224, 3) 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "bn_data (BatchNormalization)    (None, 224, 224, 3)  9           data[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d (ZeroPadding2D)  (None, 230, 230, 3)  0           bn_data[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv0 (Conv2D)                  (None, 112, 112, 64) 9408        zero_padding2d[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "bn0 (BatchNormalization)        (None, 112, 112, 64) 256         conv0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "relu0 (Activation)              (None, 112, 112, 64) 0           bn0[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_1 (ZeroPadding2D (None, 114, 114, 64) 0           relu0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "pooling0 (MaxPooling2D)         (None, 56, 56, 64)   0           zero_padding2d_1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_bn1 (BatchNormaliz (None, 56, 56, 64)   256         pooling0[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_2 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_3 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_3[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit1_sc (Conv2D)        (None, 56, 56, 64)   4096        stage1_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add (Add)                       (None, 56, 56, 64)   0           stage1_unit1_conv2[0][0]         \n",
      "                                                                 stage1_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_4 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_4[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_5 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit2_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_5[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "add_1 (Add)                     (None, 56, 56, 64)   0           stage1_unit2_conv2[0][0]         \n",
      "                                                                 add[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_relu1 (Activation) (None, 56, 56, 64)   0           stage1_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_6 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_conv1 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_6[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_bn2 (BatchNormaliz (None, 56, 56, 64)   256         stage1_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_relu2 (Activation) (None, 56, 56, 64)   0           stage1_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_7 (ZeroPadding2D (None, 58, 58, 64)   0           stage1_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage1_unit3_conv2 (Conv2D)     (None, 56, 56, 64)   36864       zero_padding2d_7[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "add_2 (Add)                     (None, 56, 56, 64)   0           stage1_unit3_conv2[0][0]         \n",
      "                                                                 add_1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_bn1 (BatchNormaliz (None, 56, 56, 64)   256         add_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_relu1 (Activation) (None, 56, 56, 64)   0           stage2_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_8 (ZeroPadding2D (None, 58, 58, 64)   0           stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_conv1 (Conv2D)     (None, 28, 28, 128)  73728       zero_padding2d_8[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_9 (ZeroPadding2D (None, 30, 30, 128)  0           stage2_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_9[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit1_sc (Conv2D)        (None, 28, 28, 128)  8192        stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_3 (Add)                     (None, 28, 28, 128)  0           stage2_unit1_conv2[0][0]         \n",
      "                                                                 stage2_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_10 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_10[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_11 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit2_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_11[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_4 (Add)                     (None, 28, 28, 128)  0           stage2_unit2_conv2[0][0]         \n",
      "                                                                 add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_12 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_12[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_13 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit3_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_13[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_5 (Add)                     (None, 28, 28, 128)  0           stage2_unit3_conv2[0][0]         \n",
      "                                                                 add_4[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_relu1 (Activation) (None, 28, 28, 128)  0           stage2_unit4_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_14 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit4_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_conv1 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_14[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_bn2 (BatchNormaliz (None, 28, 28, 128)  512         stage2_unit4_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_relu2 (Activation) (None, 28, 28, 128)  0           stage2_unit4_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_15 (ZeroPadding2 (None, 30, 30, 128)  0           stage2_unit4_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage2_unit4_conv2 (Conv2D)     (None, 28, 28, 128)  147456      zero_padding2d_15[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_6 (Add)                     (None, 28, 28, 128)  0           stage2_unit4_conv2[0][0]         \n",
      "                                                                 add_5[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_bn1 (BatchNormaliz (None, 28, 28, 128)  512         add_6[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_relu1 (Activation) (None, 28, 28, 128)  0           stage3_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_16 (ZeroPadding2 (None, 30, 30, 128)  0           stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_conv1 (Conv2D)     (None, 14, 14, 256)  294912      zero_padding2d_16[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_17 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_17[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit1_sc (Conv2D)        (None, 14, 14, 256)  32768       stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_7 (Add)                     (None, 14, 14, 256)  0           stage3_unit1_conv2[0][0]         \n",
      "                                                                 stage3_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_18 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_18[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_19 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit2_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_19[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_8 (Add)                     (None, 14, 14, 256)  0           stage3_unit2_conv2[0][0]         \n",
      "                                                                 add_7[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_20 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_20[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_21 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit3_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_21[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_9 (Add)                     (None, 14, 14, 256)  0           stage3_unit3_conv2[0][0]         \n",
      "                                                                 add_8[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit4_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_22 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit4_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_22[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit4_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit4_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_23 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit4_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit4_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_23[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_10 (Add)                    (None, 14, 14, 256)  0           stage3_unit4_conv2[0][0]         \n",
      "                                                                 add_9[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit5_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_24 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit5_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_24[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit5_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit5_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_25 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit5_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit5_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_25[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_11 (Add)                    (None, 14, 14, 256)  0           stage3_unit5_conv2[0][0]         \n",
      "                                                                 add_10[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_relu1 (Activation) (None, 14, 14, 256)  0           stage3_unit6_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_26 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit6_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_conv1 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_26[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_bn2 (BatchNormaliz (None, 14, 14, 256)  1024        stage3_unit6_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_relu2 (Activation) (None, 14, 14, 256)  0           stage3_unit6_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_27 (ZeroPadding2 (None, 16, 16, 256)  0           stage3_unit6_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage3_unit6_conv2 (Conv2D)     (None, 14, 14, 256)  589824      zero_padding2d_27[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_12 (Add)                    (None, 14, 14, 256)  0           stage3_unit6_conv2[0][0]         \n",
      "                                                                 add_11[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_bn1 (BatchNormaliz (None, 14, 14, 256)  1024        add_12[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_relu1 (Activation) (None, 14, 14, 256)  0           stage4_unit1_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_28 (ZeroPadding2 (None, 16, 16, 256)  0           stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_conv1 (Conv2D)     (None, 7, 7, 512)    1179648     zero_padding2d_28[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit1_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit1_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_29 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit1_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_29[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit1_sc (Conv2D)        (None, 7, 7, 512)    131072      stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "add_13 (Add)                    (None, 7, 7, 512)    0           stage4_unit1_conv2[0][0]         \n",
      "                                                                 stage4_unit1_sc[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_bn1 (BatchNormaliz (None, 7, 7, 512)    2048        add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_relu1 (Activation) (None, 7, 7, 512)    0           stage4_unit2_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_30 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit2_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_conv1 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_30[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit2_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit2_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_31 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit2_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit2_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_31[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_14 (Add)                    (None, 7, 7, 512)    0           stage4_unit2_conv2[0][0]         \n",
      "                                                                 add_13[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_bn1 (BatchNormaliz (None, 7, 7, 512)    2048        add_14[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_relu1 (Activation) (None, 7, 7, 512)    0           stage4_unit3_bn1[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_32 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit3_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_conv1 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_32[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_bn2 (BatchNormaliz (None, 7, 7, 512)    2048        stage4_unit3_conv1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_relu2 (Activation) (None, 7, 7, 512)    0           stage4_unit3_bn2[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "zero_padding2d_33 (ZeroPadding2 (None, 9, 9, 512)    0           stage4_unit3_relu2[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "stage4_unit3_conv2 (Conv2D)     (None, 7, 7, 512)    2359296     zero_padding2d_33[0][0]          \n",
      "__________________________________________________________________________________________________\n",
      "add_15 (Add)                    (None, 7, 7, 512)    0           stage4_unit3_conv2[0][0]         \n",
      "                                                                 add_14[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "bn1 (BatchNormalization)        (None, 7, 7, 512)    2048        add_15[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "relu1 (Activation)              (None, 7, 7, 512)    0           bn1[0][0]                        \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0_upsampling (UpSa (None, 14, 14, 512)  0           relu1[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0_concat (Concaten (None, 14, 14, 768)  0           decoder_stage0_upsampling[0][0]  \n",
      "                                                                 stage4_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_conv (Conv2D)   (None, 14, 14, 256)  1769472     decoder_stage0_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_bn (BatchNormal (None, 14, 14, 256)  1024        decoder_stage0a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0a_relu (Activatio (None, 14, 14, 256)  0           decoder_stage0a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_conv (Conv2D)   (None, 14, 14, 256)  589824      decoder_stage0a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_bn (BatchNormal (None, 14, 14, 256)  1024        decoder_stage0b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage0b_relu (Activatio (None, 14, 14, 256)  0           decoder_stage0b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1_upsampling (UpSa (None, 28, 28, 256)  0           decoder_stage0b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1_concat (Concaten (None, 28, 28, 384)  0           decoder_stage1_upsampling[0][0]  \n",
      "                                                                 stage3_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_conv (Conv2D)   (None, 28, 28, 128)  442368      decoder_stage1_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_bn (BatchNormal (None, 28, 28, 128)  512         decoder_stage1a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1a_relu (Activatio (None, 28, 28, 128)  0           decoder_stage1a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_conv (Conv2D)   (None, 28, 28, 128)  147456      decoder_stage1a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_bn (BatchNormal (None, 28, 28, 128)  512         decoder_stage1b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage1b_relu (Activatio (None, 28, 28, 128)  0           decoder_stage1b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2_upsampling (UpSa (None, 56, 56, 128)  0           decoder_stage1b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2_concat (Concaten (None, 56, 56, 192)  0           decoder_stage2_upsampling[0][0]  \n",
      "                                                                 stage2_unit1_relu1[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_conv (Conv2D)   (None, 56, 56, 64)   110592      decoder_stage2_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_bn (BatchNormal (None, 56, 56, 64)   256         decoder_stage2a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2a_relu (Activatio (None, 56, 56, 64)   0           decoder_stage2a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_conv (Conv2D)   (None, 56, 56, 64)   36864       decoder_stage2a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_bn (BatchNormal (None, 56, 56, 64)   256         decoder_stage2b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage2b_relu (Activatio (None, 56, 56, 64)   0           decoder_stage2b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3_upsampling (UpSa (None, 112, 112, 64) 0           decoder_stage2b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3_concat (Concaten (None, 112, 112, 128 0           decoder_stage3_upsampling[0][0]  \n",
      "                                                                 relu0[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_conv (Conv2D)   (None, 112, 112, 32) 36864       decoder_stage3_concat[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_bn (BatchNormal (None, 112, 112, 32) 128         decoder_stage3a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3a_relu (Activatio (None, 112, 112, 32) 0           decoder_stage3a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_conv (Conv2D)   (None, 112, 112, 32) 9216        decoder_stage3a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_bn (BatchNormal (None, 112, 112, 32) 128         decoder_stage3b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage3b_relu (Activatio (None, 112, 112, 32) 0           decoder_stage3b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4_upsampling (UpSa (None, 224, 224, 32) 0           decoder_stage3b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_conv (Conv2D)   (None, 224, 224, 16) 4608        decoder_stage4_upsampling[0][0]  \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_bn (BatchNormal (None, 224, 224, 16) 64          decoder_stage4a_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4a_relu (Activatio (None, 224, 224, 16) 0           decoder_stage4a_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_conv (Conv2D)   (None, 224, 224, 16) 2304        decoder_stage4a_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_bn (BatchNormal (None, 224, 224, 16) 64          decoder_stage4b_conv[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "decoder_stage4b_relu (Activatio (None, 224, 224, 16) 0           decoder_stage4b_bn[0][0]         \n",
      "__________________________________________________________________________________________________\n",
      "final_conv (Conv2D)             (None, 224, 224, 2)  290         decoder_stage4b_relu[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "softmax (Activation)            (None, 224, 224, 2)  0           final_conv[0][0]                 \n",
      "==================================================================================================\n",
      "Total params: 24,456,299\n",
      "Trainable params: 3,167,205\n",
      "Non-trainable params: 21,289,094\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = tf.keras.models.load_model('resnet_norm_finetune.hdf5')\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = model.predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((8382, 224, 224, 2), (8382, 224, 224, 2))"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.shape, Y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_iou(gt, pred):\n",
    "    return (gt & pred).sum() / (gt | pred).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9511608450115039\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "i = 0\n",
    "visualize(Y[i][:, :, 1:]> 0.5, (y[i][:, :, 1:] > 0.5), X[i].astype('float32'))\n",
    "print(get_iou((Y[i][:, :, 1:]> 0.5), (y[i][:, :, 1:] > 0.5)))\n",
    "#plt.imshow(X[i].astype('float32'))\n",
    "#plt.imshow((Y[i][:, :, 1:] > 0.5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "105 0.49274809160305344\n",
      "155 0.48473967684021546\n",
      "158 0.360149236329719\n",
      "183 0.32834487379941923\n",
      "194 0.28771929824561404\n",
      "249 0.3155053358443189\n",
      "268 0.23784254023488474\n",
      "321 0.25071664829106943\n",
      "383 0.43903940886699505\n",
      "475 0.48094877617966186\n",
      "604 0.4788285109386027\n",
      "755 0.43307314252674184\n",
      "839 0.4149494556765163\n",
      "873 0.25851293103448275\n",
      "875 0.12624233453161346\n",
      "1235 0.3888583470771486\n",
      "1330 0.4792105263157895\n",
      "1502 0.35195416410886027\n",
      "1612 0.20112907855707587\n",
      "2111 0.45419187554019014\n",
      "2135 0.3160044559970293\n",
      "2171 0.4217718059667502\n",
      "2183 0.43729754743174454\n",
      "2401 0.4270907791279485\n",
      "2686 0.3641940881204685\n",
      "2779 0.21103217718345174\n",
      "2796 0.20721311475409837\n",
      "2811 0.42057761732851984\n",
      "3409 0.0\n",
      "3577 0.32061440936242835\n",
      "3812 0.3982311755312615\n",
      "4176 0.4573444515715202\n",
      "4304 0.045898852528686786\n",
      "4340 0.362791658063184\n",
      "4345 0.44107821679530007\n",
      "4784 0.0\n",
      "4990 0.440625\n",
      "5139 0.2540216984661429\n",
      "5228 0.48511761881901105\n",
      "5511 0.24775433465636099\n",
      "5614 0.06423982869379015\n",
      "5689 0.47584233947870314\n",
      "5740 0.499911987326175\n",
      "5958 0.49050279329608937\n",
      "6186 0.37414187643020597\n",
      "6939 0.17173837005946135\n",
      "7203 0.49757372153788726\n",
      "7231 0.4518593988792664\n",
      "7626 0.017512746619374863\n",
      "7658 0.3484848484848485\n",
      "7798 0.25019384853967436\n",
      "7858 0.3465264323277055\n",
      "7939 0.3429547395388557\n",
      "8028 0.4868217054263566\n",
      "0.8787639104961682\n"
     ]
    }
   ],
   "source": [
    "arr = []\n",
    "for i in range(8382):\n",
    "    #i = 34\n",
    "    #visualize(Y_test[i][:, :, 1:].astype(int), (y[i][:, :, 1:] > 0.5))\n",
    "    iou = get_iou(Y[i][:, :, 1:] > 0.5, (y[i][:, :, 1:] > 0.5))\n",
    "    arr.append(iou)\n",
    "    if iou < 0.5:\n",
    "        print(i, iou)\n",
    "    #print(arr[-1])\n",
    "print(np.mean(arr))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}