vgg16_mergemodel.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import glob\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from tensorflow.keras.layers import Flatten, Input, Dense, BatchNormalization, Conv2D, MaxPool2D, GlobalMaxPool2D, GlobalAveragePooling2D, Dropout, Add\n",
    "from tensorflow.keras.optimizers import SGD\n",
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.applications.vgg16 import VGG16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_11 (InputLayer)           (None, None, None, 3 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "block1_conv1 (Conv2D)           (None, None, None, 6 1792        input_11[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "block1_conv2 (Conv2D)           (None, None, None, 6 36928       block1_conv1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block1_pool (MaxPooling2D)      (None, None, None, 6 0           block1_conv2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block2_conv1 (Conv2D)           (None, None, None, 1 73856       block1_pool[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "block2_conv2 (Conv2D)           (None, None, None, 1 147584      block2_conv1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block2_pool (MaxPooling2D)      (None, None, None, 1 0           block2_conv2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block3_conv1 (Conv2D)           (None, None, None, 2 295168      block2_pool[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "block3_conv2 (Conv2D)           (None, None, None, 2 590080      block3_conv1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block3_conv3 (Conv2D)           (None, None, None, 2 590080      block3_conv2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block3_pool (MaxPooling2D)      (None, None, None, 2 0           block3_conv3[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block4_conv1 (Conv2D)           (None, None, None, 5 1180160     block3_pool[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "block4_conv2 (Conv2D)           (None, None, None, 5 2359808     block4_conv1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block4_conv3 (Conv2D)           (None, None, None, 5 2359808     block4_conv2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block4_pool (MaxPooling2D)      (None, None, None, 5 0           block4_conv3[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block5_conv1 (Conv2D)           (None, None, None, 5 2359808     block4_pool[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "block5_conv2 (Conv2D)           (None, None, None, 5 2359808     block5_conv1[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block5_conv3 (Conv2D)           (None, None, None, 5 2359808     block5_conv2[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "block5_pool (MaxPooling2D)      (None, None, None, 5 0           block5_conv3[0][0]               \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_3 (Glo (None, 512)          0           block5_pool[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "flatten_3 (Flatten)             (None, 512)          0           global_average_pooling2d_3[0][0] \n",
      "__________________________________________________________________________________________________\n",
      "input_12 (InputLayer)           (None, None, 1)      0                                            \n",
      "__________________________________________________________________________________________________\n",
      "add_3 (Add)                     (None, None, 512)    0           flatten_3[0][0]                  \n",
      "                                                                 input_12[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_12 (Dense)                (None, None, 1024)   525312      add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "dense_15 (Dense)                (None, None, 1024)   525312      add_3[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "dense_13 (Dense)                (None, None, 512)    524800      dense_12[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_16 (Dense)                (None, None, 512)    524800      dense_15[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_14 (Dense)                (None, None, 128)    65664       dense_13[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_17 (Dense)                (None, None, 128)    65664       dense_16[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "left_length (Dense)             (None, None, 1)      129         dense_14[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "right_length (Dense)            (None, None, 1)      129         dense_17[0][0]                   \n",
      "==================================================================================================\n",
      "Total params: 16,946,498\n",
      "Trainable params: 2,231,810\n",
      "Non-trainable params: 14,714,688\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "input1 = Input(shape=(None, None, 3))\n",
    "input2 = Input(shape=(None,1))\n",
    "\n",
    "base_model = VGG16(include_top=False,input_tensor = input1)\n",
    "x = base_model.output\n",
    "x = GlobalAveragePooling2D()(x)\n",
    "x = Flatten()(x)\n",
    "\n",
    "merge = Add()([x,input2])\n",
    "\n",
    "left = Dense(units=1024,activation='relu')(merge)\n",
    "left = Dense(units=512,activation='relu')(left)\n",
    "left = Dense(units=128,activation='relu')(left)\n",
    "\n",
    "left_length = Dense(units=1,activation='linear',name='left_length')(left)\n",
    "\n",
    "right = Dense(units=1024,activation='relu')(merge)\n",
    "right = Dense(units=512,activation='relu')(right)\n",
    "right = Dense(units=128,activation='relu')(right)\n",
    "\n",
    "right_length = Dense(units=1,activation='linear',name='right_length')(right)\n",
    "\n",
    "model = Model(inputs=[input1,input2],outputs=[left_length,right_length])\n",
    "for layer in base_model.layers:\n",
    "    layer.trainable = False\n",
    "\n",
    "model.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "name": "pycharm-37187e8b",
   "language": "python",
   "display_name": "PyCharm (handson-ml2)"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}