Alexnet on TESS dataset.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: opencv-python in c:\\users\\shireen\\lib\\site-packages (4.4.0.40)\n",
      "Requirement already satisfied: numpy>=1.17.3 in c:\\users\\shireen\\lib\\site-packages (from opencv-python) (1.18.5)\n"
     ]
    }
   ],
   "source": [
    "!pip install opencv-python\n",
    "import os\n",
    "import cv2\n",
    "import numpy as np\n",
    "from PIL import Image\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No. of images loaded =  2800 \n",
      "No. of labels loaded =  2800\n",
      "<class 'list'> <class 'list'>\n"
     ]
    }
   ],
   "source": [
    "categories = os.listdir(\"C:/Users/Shireen/A FINAL TESS ATTEMPT/data/spectograms/\")\n",
    "def load_images_and_labels(categories):\n",
    "    img_lst=[]\n",
    "    labels=[]\n",
    "    for index, category in enumerate(categories):\n",
    "        for image_name in os.listdir(fpath+\"/\"+category):\n",
    "            img = cv2.imread(fpath+\"/\"+category+\"/\"+image_name)\n",
    "            img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
    "            \n",
    "            img_array = Image.fromarray(img, 'RGB')\n",
    "            \n",
    "            #resize image to 227 x 227 because the input image resolution for Alexnet is 227 x 227\n",
    "            resized_img = img_array.resize((227, 227))\n",
    "            \n",
    "            img_lst.append(np.array(resized_img))\n",
    "            \n",
    "            labels.append(index)\n",
    "    return img_lst, labels\n",
    "\n",
    "fpath = \"C:/Users/Shireen/A FINAL TESS ATTEMPT/data/spectograms/\"\n",
    "images, labels = load_images_and_labels(categories)\n",
    "print(\"No. of images loaded = \",len(images),\"\\nNo. of labels loaded = \",len(labels))\n",
    "print(type(images),type(labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Images shape =  (2800, 227, 227, 3) \n",
      "Labels shape =  (2800,)\n",
      "<class 'numpy.ndarray'> <class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "images = np.array(images)\n",
    "labels = np.array(labels)\n",
    "\n",
    "print(\"Images shape = \",images.shape,\"\\nLabels shape = \",labels.shape)\n",
    "print(type(images),type(labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "'n' values before shuffling =  [   0    1    2 ... 2797 2798 2799]\n",
      "\n",
      "'n' values after shuffling =  [1088  772 2161 ... 1130 1294  860]\n"
     ]
    }
   ],
   "source": [
    "#1-step in data shuffling\n",
    "random_seed=42\n",
    "#get equally spaced numbers in a given range\n",
    "n = np.arange(images.shape[0])\n",
    "print(\"'n' values before shuffling = \",n)\n",
    "\n",
    "#shuffle all the equally spaced values in list 'n'\n",
    "np.random.seed(random_seed)\n",
    "np.random.shuffle(n)\n",
    "print(\"\\n'n' values after shuffling = \",n)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Images shape after shuffling =  (2800, 227, 227, 3) \n",
      "Labels shape after shuffling =  (2800,)\n"
     ]
    }
   ],
   "source": [
    "#2-step in data shuffling\n",
    "\n",
    "#shuffle images and corresponding labels data in both the lists\n",
    "images = images[n]\n",
    "labels = labels[n]\n",
    "\n",
    "print(\"Images shape after shuffling = \",images.shape,\"\\nLabels shape after shuffling = \",labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Images shape after normalization =  (2800, 227, 227, 3)\n"
     ]
    }
   ],
   "source": [
    "images = images.astype(np.float32)\n",
    "labels = labels.astype(np.int32)\n",
    "images = images/255\n",
    "print(\"Images shape after normalization = \",images.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train shape =  (2240, 227, 227, 3)\n",
      "y_train shape =  (2240,)\n",
      "\n",
      "x_test shape =  (560, 227, 227, 3)\n",
      "y_test shape =  (560,)\n"
     ]
    }
   ],
   "source": [
    "x_train, x_test, y_train, y_test = train_test_split(images, labels, test_size = 0.2, random_state = random_seed)\n",
    "\n",
    "print(\"x_train shape = \",x_train.shape)\n",
    "print(\"y_train shape = \",y_train.shape)\n",
    "print(\"\\nx_test shape = \",x_test.shape)\n",
    "print(\"y_test shape = \",y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D,MaxPooling2D,Dense,Flatten,Dropout\n",
    "from keras.layers.normalization import BatchNormalization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_5 (Conv2D)            (None, 55, 55, 96)        34944     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 27, 27, 96)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_6 (Batch (None, 27, 27, 96)        384       \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 23, 23, 256)       614656    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 11, 11, 256)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_7 (Batch (None, 11, 11, 256)       1024      \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 9, 9, 384)         885120    \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 7, 7, 384)         1327488   \n",
      "_________________________________________________________________\n",
      "conv2d_9 (Conv2D)            (None, 5, 5, 256)         884992    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 2, 2, 256)         0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_8 (Batch (None, 2, 2, 256)         1024      \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 4096)              4198400   \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 4096)              0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_9 (Batch (None, 4096)              16384     \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 4096)              16781312  \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 4096)              0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_10 (Batc (None, 4096)              16384     \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 1000)              4097000   \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 1000)              0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_11 (Batc (None, 1000)              4000      \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 20)                20020     \n",
      "=================================================================\n",
      "Total params: 28,883,132\n",
      "Trainable params: 28,863,532\n",
      "Non-trainable params: 19,600\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model=Sequential()\n",
    "\n",
    "#1 conv layer\n",
    "model.add(Conv2D(filters=96,kernel_size=(11,11),strides=(4,4),padding=\"valid\",activation=\"relu\",input_shape=(227,227,3)))\n",
    "\n",
    "#1 max pool layer\n",
    "model.add(MaxPooling2D(pool_size=(3,3),strides=(2,2)))\n",
    "\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "#2 conv layer\n",
    "model.add(Conv2D(filters=256,kernel_size=(5,5),strides=(1,1),padding=\"valid\",activation=\"relu\"))\n",
    "\n",
    "#2 max pool layer\n",
    "model.add(MaxPooling2D(pool_size=(3,3),strides=(2,2)))\n",
    "\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "#3 conv layer\n",
    "model.add(Conv2D(filters=384,kernel_size=(3,3),strides=(1,1),padding=\"valid\",activation=\"relu\"))\n",
    "\n",
    "#4 conv layer\n",
    "model.add(Conv2D(filters=384,kernel_size=(3,3),strides=(1,1),padding=\"valid\",activation=\"relu\"))\n",
    "\n",
    "#5 conv layer\n",
    "model.add(Conv2D(filters=256,kernel_size=(3,3),strides=(1,1),padding=\"valid\",activation=\"relu\"))\n",
    "\n",
    "#3 max pool layer\n",
    "model.add(MaxPooling2D(pool_size=(3,3),strides=(2,2)))\n",
    "\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "\n",
    "model.add(Flatten())\n",
    "\n",
    "#1 dense layer\n",
    "model.add(Dense(4096,input_shape=(227,227,3),activation=\"relu\"))\n",
    "\n",
    "model.add(Dropout(0.4))\n",
    "\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "#2 dense layer\n",
    "model.add(Dense(4096,activation=\"relu\"))\n",
    "\n",
    "model.add(Dropout(0.4))\n",
    "\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "#3 dense layer\n",
    "model.add(Dense(1000,activation=\"relu\"))\n",
    "\n",
    "model.add(Dropout(0.4))\n",
    "\n",
    "model.add(BatchNormalization())\n",
    "\n",
    "#output layer\n",
    "model.add(Dense(20,activation=\"softmax\"))\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer=\"adam\", loss=\"sparse_categorical_crossentropy\", metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.0729 - accuracy: 0.7183\n",
      "Epoch 2/20\n",
      "18/18 [==============================] - 123s 7s/step - loss: 0.1304 - accuracy: 0.9607\n",
      "Epoch 3/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0589 - accuracy: 0.9812\n",
      "Epoch 4/20\n",
      "18/18 [==============================] - 122s 7s/step - loss: 0.0349 - accuracy: 0.9911\n",
      "Epoch 5/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0167 - accuracy: 0.9937\n",
      "Epoch 6/20\n",
      "18/18 [==============================] - 122s 7s/step - loss: 0.0101 - accuracy: 0.9973\n",
      "Epoch 7/20\n",
      "18/18 [==============================] - 122s 7s/step - loss: 0.0294 - accuracy: 0.9929\n",
      "Epoch 8/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0098 - accuracy: 0.9964\n",
      "Epoch 9/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0109 - accuracy: 0.9960\n",
      "Epoch 10/20\n",
      "18/18 [==============================] - 122s 7s/step - loss: 0.0199 - accuracy: 0.9951\n",
      "Epoch 11/20\n",
      "18/18 [==============================] - 123s 7s/step - loss: 0.0061 - accuracy: 0.9987\n",
      "Epoch 12/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0041 - accuracy: 0.9991\n",
      "Epoch 13/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0132 - accuracy: 0.9964\n",
      "Epoch 14/20\n",
      "18/18 [==============================] - 127s 7s/step - loss: 0.0289 - accuracy: 0.9929\n",
      "Epoch 15/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0077 - accuracy: 0.9978\n",
      "Epoch 16/20\n",
      "18/18 [==============================] - 122s 7s/step - loss: 0.0068 - accuracy: 0.9978\n",
      "Epoch 17/20\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.0062 - accuracy: 0.9978\n",
      "Epoch 18/20\n",
      "18/18 [==============================] - 123s 7s/step - loss: 0.0029 - accuracy: 0.9996\n",
      "Epoch 19/20\n",
      "18/18 [==============================] - 130s 7s/step - loss: 0.0014 - accuracy: 0.9996\n",
      "Epoch 20/20\n",
      "18/18 [==============================] - 125s 7s/step - loss: 0.0045 - accuracy: 0.9991\n",
      "Wall time: 43min 23s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x1edc2681940>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "model.fit(x_train, y_train, epochs=10,batch_size=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 0.7960949540138245\n",
      "Test accuracy: 0.8232142925262451\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import Dense, Flatten, Reshape, Input, InputLayer\n",
    "from keras.models import Sequential, Model\n",
    "\n",
    "def build_autoencoder(img_shape, code_size):\n",
    "    # The encoder\n",
    "    encoder = Sequential()\n",
    "    encoder.add(InputLayer(img_shape))\n",
    "    encoder.add(Flatten())\n",
    "    encoder.add(Dense(code_size))\n",
    "\n",
    "    # The decoder\n",
    "    decoder = Sequential()\n",
    "    decoder.add(InputLayer((code_size,)))\n",
    "    decoder.add(Dense(np.prod(img_shape)))\n",
    "    decoder.add(Reshape(img_shape))\n",
    "\n",
    "    return encoder, decoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"functional_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_3 (InputLayer)         [(None, 227, 227, 3)]     0         \n",
      "_________________________________________________________________\n",
      "sequential_1 (Sequential)    (None, 32)                4946816   \n",
      "_________________________________________________________________\n",
      "sequential_2 (Sequential)    (None, 227, 227, 3)       5101371   \n",
      "=================================================================\n",
      "Total params: 10,048,187\n",
      "Trainable params: 10,048,187\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "IMG_SHAPE = images.shape[1:]\n",
    "encoder, decoder = build_autoencoder(IMG_SHAPE, 32)\n",
    "\n",
    "inp = Input(IMG_SHAPE)\n",
    "code = encoder(inp)\n",
    "reconstruction = decoder(code)\n",
    "\n",
    "autoencoder = Model(inp,reconstruction)\n",
    "autoencoder.compile(optimizer='adamax', loss='mse')\n",
    "\n",
    "print(autoencoder.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "70/70 [==============================] - 10s 136ms/step - loss: 0.3361 - val_loss: 0.0000e+00\n",
      "Epoch 2/20\n",
      "70/70 [==============================] - 9s 135ms/step - loss: 0.0146 - val_loss: 0.0000e+00\n",
      "Epoch 3/20\n",
      "70/70 [==============================] - 10s 137ms/step - loss: 0.0144 - val_loss: 0.0000e+00\n",
      "Epoch 4/20\n",
      "70/70 [==============================] - 10s 136ms/step - loss: 0.0144 - val_loss: 0.0000e+00\n",
      "Epoch 5/20\n",
      "70/70 [==============================] - 10s 137ms/step - loss: 0.0144 - val_loss: 0.0000e+00\n",
      "Epoch 6/20\n",
      "70/70 [==============================] - 10s 138ms/step - loss: 0.0143 - val_loss: 0.0000e+00\n",
      "Epoch 7/20\n",
      "70/70 [==============================] - 10s 137ms/step - loss: 0.0143 - val_loss: 0.0000e+00\n",
      "Epoch 8/20\n",
      "70/70 [==============================] - 10s 138ms/step - loss: 0.0143 - val_loss: 0.0000e+00\n",
      "Epoch 9/20\n",
      "70/70 [==============================] - 10s 138ms/step - loss: 0.0143 - val_loss: 0.0000e+00\n",
      "Epoch 10/20\n",
      "70/70 [==============================] - 10s 139ms/step - loss: 0.0143 - val_loss: 0.0000e+00\n",
      "Epoch 11/20\n",
      "70/70 [==============================] - 10s 139ms/step - loss: 0.0142 - val_loss: 0.0000e+00\n",
      "Epoch 12/20\n",
      "70/70 [==============================] - 10s 139ms/step - loss: 0.0142 - val_loss: 0.0000e+00\n",
      "Epoch 13/20\n",
      "70/70 [==============================] - 10s 138ms/step - loss: 0.0142 - val_loss: 0.0000e+00\n",
      "Epoch 14/20\n",
      "70/70 [==============================] - 11s 154ms/step - loss: 0.0142 - val_loss: 0.0000e+00\n",
      "Epoch 15/20\n",
      "70/70 [==============================] - 12s 164ms/step - loss: 0.0141 - val_loss: 0.0000e+00\n",
      "Epoch 16/20\n",
      "70/70 [==============================] - 10s 136ms/step - loss: 0.0141 - val_loss: 0.0000e+00\n",
      "Epoch 17/20\n",
      "70/70 [==============================] - 9s 135ms/step - loss: 0.0140 - val_loss: 0.0000e+00\n",
      "Epoch 18/20\n",
      "70/70 [==============================] - 9s 133ms/step - loss: 0.0139 - val_loss: 0.0000e+00\n",
      "Epoch 19/20\n",
      "70/70 [==============================] - 9s 134ms/step - loss: 0.0137 - val_loss: 0.0000e+00\n",
      "Epoch 20/20\n",
      "70/70 [==============================] - 10s 138ms/step - loss: 0.0136 - val_loss: 0.0000e+00\n"
     ]
    }
   ],
   "source": [
    "history = autoencoder.fit(x=x_train, y=x_train, epochs=20,\n",
    "                validation_data=[x_test, x_test])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy= 0.9316340610384941 \n",
      "\n"
     ]
    }
   ],
   "source": [
    "decoded_imgs = autoencoder.predict(images)\n",
    "\n",
    "n = 1\n",
    "plt.figure(figsize=(20, 4))\n",
    "for i in range(n):\n",
    "    # display original\n",
    "    ax = plt.subplot(2, n, i+1)\n",
    "    plt.imshow(images[i].reshape(227, 227,3))\n",
    "    plt.gray()\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "\n",
    "    # display reconstruction\n",
    "    ax = plt.subplot(2, n, i + n+1)\n",
    "    plt.imshow(decoded_imgs[i].reshape(227, 227,3))\n",
    "    plt.gray()\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "plt.show()\n",
    "\n",
    "print(\"Accuracy=\",1-np.mean(abs(images-decoded_imgs)),'\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x_train shape =  (2240, 227, 227, 3)\n",
      "y_train shape =  (2240,)\n",
      "\n",
      "x_test shape =  (560, 227, 227, 3)\n",
      "y_test shape =  (560,)\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, Y_train, Y_test = train_test_split(decoded_imgs, labels, test_size = 0.2, random_state = random_seed)\n",
    "\n",
    "print(\"x_train shape = \",X_train.shape)\n",
    "print(\"y_train shape = \",Y_train.shape)\n",
    "print(\"\\nx_test shape = \",X_test.shape)\n",
    "print(\"y_test shape = \",Y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 1.5275 - accuracy: 0.4165\n",
      "Epoch 2/100\n",
      "18/18 [==============================] - 123s 7s/step - loss: 1.5510 - accuracy: 0.3991\n",
      "Epoch 3/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.5077 - accuracy: 0.4259\n",
      "Epoch 4/100\n",
      "18/18 [==============================] - 119s 7s/step - loss: 1.4886 - accuracy: 0.4223\n",
      "Epoch 5/100\n",
      "18/18 [==============================] - 119s 7s/step - loss: 1.4838 - accuracy: 0.4286\n",
      "Epoch 6/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.4192 - accuracy: 0.4460\n",
      "Epoch 7/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 1.3833 - accuracy: 0.4589\n",
      "Epoch 8/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 1.3664 - accuracy: 0.4714\n",
      "Epoch 9/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 1.3301 - accuracy: 0.4821\n",
      "Epoch 10/100\n",
      "18/18 [==============================] - 132s 7s/step - loss: 1.3461 - accuracy: 0.4759\n",
      "Epoch 11/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.3150 - accuracy: 0.4875\n",
      "Epoch 12/100\n",
      "18/18 [==============================] - 124s 7s/step - loss: 1.2663 - accuracy: 0.5094\n",
      "Epoch 13/100\n",
      "18/18 [==============================] - 123s 7s/step - loss: 1.2484 - accuracy: 0.5085\n",
      "Epoch 14/100\n",
      "18/18 [==============================] - 123s 7s/step - loss: 1.2546 - accuracy: 0.5058\n",
      "Epoch 15/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.3295 - accuracy: 0.4795\n",
      "Epoch 16/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.2934 - accuracy: 0.4897\n",
      "Epoch 17/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.2357 - accuracy: 0.5241\n",
      "Epoch 18/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 1.2345 - accuracy: 0.5179\n",
      "Epoch 19/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.1943 - accuracy: 0.5353\n",
      "Epoch 20/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 1.1980 - accuracy: 0.5397\n",
      "Epoch 21/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.1614 - accuracy: 0.5482\n",
      "Epoch 22/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 1.1915 - accuracy: 0.5353\n",
      "Epoch 23/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.1982 - accuracy: 0.5192\n",
      "Epoch 24/100\n",
      "18/18 [==============================] - 125s 7s/step - loss: 1.1572 - accuracy: 0.5509\n",
      "Epoch 25/100\n",
      "18/18 [==============================] - 124s 7s/step - loss: 1.1622 - accuracy: 0.5500\n",
      "Epoch 26/100\n",
      "18/18 [==============================] - 124s 7s/step - loss: 1.1145 - accuracy: 0.5594\n",
      "Epoch 27/100\n",
      "18/18 [==============================] - 123s 7s/step - loss: 1.0752 - accuracy: 0.5777\n",
      "Epoch 28/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.0987 - accuracy: 0.5652\n",
      "Epoch 29/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.1282 - accuracy: 0.5603\n",
      "Epoch 30/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.0664 - accuracy: 0.5884\n",
      "Epoch 31/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.0520 - accuracy: 0.5911\n",
      "Epoch 32/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.0917 - accuracy: 0.5835\n",
      "Epoch 33/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.0726 - accuracy: 0.5768\n",
      "Epoch 34/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 1.0192 - accuracy: 0.5946\n",
      "Epoch 35/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.9879 - accuracy: 0.6062\n",
      "Epoch 36/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.9714 - accuracy: 0.6250\n",
      "Epoch 37/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.9486 - accuracy: 0.6357\n",
      "Epoch 38/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 1.0416 - accuracy: 0.6004\n",
      "Epoch 39/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.9986 - accuracy: 0.6205\n",
      "Epoch 40/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.9570 - accuracy: 0.6295\n",
      "Epoch 41/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.9205 - accuracy: 0.6348\n",
      "Epoch 42/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.9198 - accuracy: 0.6446\n",
      "Epoch 43/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8990 - accuracy: 0.6603\n",
      "Epoch 44/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.8646 - accuracy: 0.6683\n",
      "Epoch 45/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8471 - accuracy: 0.6777\n",
      "Epoch 46/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8210 - accuracy: 0.6795\n",
      "Epoch 47/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8527 - accuracy: 0.6710\n",
      "Epoch 48/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8459 - accuracy: 0.6897\n",
      "Epoch 49/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8209 - accuracy: 0.6853\n",
      "Epoch 50/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8147 - accuracy: 0.6844\n",
      "Epoch 51/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8012 - accuracy: 0.6893\n",
      "Epoch 52/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.8221 - accuracy: 0.6884\n",
      "Epoch 53/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8207 - accuracy: 0.6857\n",
      "Epoch 54/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.7896 - accuracy: 0.6996\n",
      "Epoch 55/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.7772 - accuracy: 0.6996\n",
      "Epoch 56/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.7614 - accuracy: 0.7112\n",
      "Epoch 57/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.8268 - accuracy: 0.6839\n",
      "Epoch 58/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.7503 - accuracy: 0.7165\n",
      "Epoch 59/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.7338 - accuracy: 0.7094\n",
      "Epoch 60/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.7424 - accuracy: 0.7179\n",
      "Epoch 61/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.8070 - accuracy: 0.6915\n",
      "Epoch 62/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.7345 - accuracy: 0.7174\n",
      "Epoch 63/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6944 - accuracy: 0.7290\n",
      "Epoch 64/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6651 - accuracy: 0.7487\n",
      "Epoch 65/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.7637 - accuracy: 0.7143\n",
      "Epoch 66/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.6765 - accuracy: 0.7420\n",
      "Epoch 67/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.6606 - accuracy: 0.7504\n",
      "Epoch 68/100\n",
      "18/18 [==============================] - 123s 7s/step - loss: 0.6604 - accuracy: 0.7429\n",
      "Epoch 69/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6630 - accuracy: 0.7411\n",
      "Epoch 70/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.6502 - accuracy: 0.7482\n",
      "Epoch 71/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6432 - accuracy: 0.7549\n",
      "Epoch 72/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6371 - accuracy: 0.7536\n",
      "Epoch 73/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.6835 - accuracy: 0.7330\n",
      "Epoch 74/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6786 - accuracy: 0.7254\n",
      "Epoch 75/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6733 - accuracy: 0.7451\n",
      "Epoch 76/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6435 - accuracy: 0.7487\n",
      "Epoch 77/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6025 - accuracy: 0.7799\n",
      "Epoch 78/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6027 - accuracy: 0.7665\n",
      "Epoch 79/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.5917 - accuracy: 0.7705\n",
      "Epoch 80/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.6044 - accuracy: 0.7696\n",
      "Epoch 81/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.5780 - accuracy: 0.7862\n",
      "Epoch 82/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.6410 - accuracy: 0.7487\n",
      "Epoch 83/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.5744 - accuracy: 0.7830\n",
      "Epoch 84/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.5666 - accuracy: 0.7812\n",
      "Epoch 85/100\n",
      "18/18 [==============================] - 123s 7s/step - loss: 0.5795 - accuracy: 0.7812\n",
      "Epoch 86/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.6013 - accuracy: 0.7781\n",
      "Epoch 87/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 0.5559 - accuracy: 0.7848\n",
      "Epoch 88/100\n",
      "18/18 [==============================] - 122s 7s/step - loss: 0.5464 - accuracy: 0.7929\n",
      "Epoch 89/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.5101 - accuracy: 0.8103\n",
      "Epoch 90/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.5099 - accuracy: 0.8076\n",
      "Epoch 91/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.5161 - accuracy: 0.8049\n",
      "Epoch 92/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.5552 - accuracy: 0.7982\n",
      "Epoch 93/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.5264 - accuracy: 0.8058\n",
      "Epoch 94/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.5140 - accuracy: 0.7987\n",
      "Epoch 95/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.4942 - accuracy: 0.8179\n",
      "Epoch 96/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.4929 - accuracy: 0.8170\n",
      "Epoch 97/100\n",
      "18/18 [==============================] - 120s 7s/step - loss: 0.4957 - accuracy: 0.8134\n",
      "Epoch 98/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.4884 - accuracy: 0.8147\n",
      "Epoch 99/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.5050 - accuracy: 0.8125\n",
      "Epoch 100/100\n",
      "18/18 [==============================] - 121s 7s/step - loss: 0.4885 - accuracy: 0.8125\n",
      "Wall time: 3h 33min 49s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x1edbf721910>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "model.fit(X_train, Y_train, epochs=100, batch_size=128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test loss: 3.35880446434021\n",
      "Test accuracy: 0.4303571283817291\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate(X_test, Y_test, verbose=0)\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy:', score[1])"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}