D_Feature_Extraction_VGG16_1.ipynb

{
 "cells": [
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "Entreno de un ConvNet VGG16 preentrenado con Imagenet y\n",
    "entrenando conjunto de datos pequeño\n",
    "Conjunto de datos de 4 clases\n",
    "Utilizaremos un conjunto de datos reducido:\n",
    "- Conjunto de entrenamiento: 1.000 muestras x clase \n",
    "- Conjunto de validación: 300 muestras x clase\n",
    "- Conjunto de test: 300 muestras x clase\n",
    "El conjunto de datos: balanced_dataset\n",
    "contiene tres carpetas, train, validation y test, cada una con 4 clases"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from  keras.preprocessing.image  import ImageDataGenerator, img_to_array, load_img \n",
    "from  keras.models  import Sequential\n",
    "from  keras.layers  import Dropout,  Flatten,  Dense\n",
    "from  keras.applications.vgg16 import VGG16 \n",
    "from keras.optimizers import RMSprop\n",
    "from  keras.utils.np_utils import to_categorical \n",
    "import time\n",
    "import os\n",
    "import numpy as np \n",
    "import math\n",
    "import  matplotlib.pyplot  as  plt\n",
    "\n",
    "img_width = 224\n",
    "img_height = 224\n",
    "batch_size = 64 \n",
    "epochs = 80\n",
    "nb_train = 4000\n",
    "nb_validation = 1200\n",
    "nb_test = 1200\n",
    "nb_classes = 4\n",
    "\n",
    "PATH_TO_FE = \"VGG\" \n",
    "train_dir  =  \"balanced_dataset/train\"\n",
    "validation_dir = \"balanced_dataset/validation\" \n",
    "test_dir  =  \"balanced_dataset/test\" \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Load Pretrained model\n",
    "model = VGG16(include_top=False,  weights='imagenet') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 4000 images belonging to 4 classes.\n",
      "Found 1200 images belonging to 4 classes.\n",
      "Found 1200 images belonging to 4 classes.\n"
     ]
    }
   ],
   "source": [
    "#PREPROCESSING + FEATURE EXTRACTION\n",
    "\n",
    "datagen  =  ImageDataGenerator(rescale=1. / 255)\n",
    "\n",
    "train_generator = datagen.flow_from_directory( \n",
    "    train_dir,\n",
    "    target_size = (img_width, img_height), \n",
    "    batch_size = batch_size,\n",
    "    class_mode  =  None,  \n",
    "    shuffle = False, \n",
    "    interpolation  =  'lanczos')\n",
    "\n",
    "validation_generator = datagen.flow_from_directory( \n",
    "    validation_dir, \n",
    "    target_size=(img_width, img_height), \n",
    "    batch_size=batch_size, \n",
    "    class_mode=None,\n",
    "    shuffle=False, \n",
    "    interpolation  =  'lanczos')\n",
    "\n",
    "test_generator = datagen.flow_from_directory(\n",
    "    test_dir,\n",
    "    target_size = (img_width, img_height),\n",
    "    batch_size = batch_size,\n",
    "    class_mode = None,\n",
    "    shuffle=False, \n",
    "    interpolation  =  'lanczos')\n",
    "\n",
    "###########Feature Extraction (FE)#########################\n",
    "#Train\n",
    "max_size_train = int(math.ceil(nb_train / batch_size)) \n",
    "train_features = model.predict_generator(train_generator, max_size_train)\n",
    "np.save('VGG/VGG16_FE_train_1.npy',  train_features)\n",
    "#Validation\n",
    "max_size_validation = int(math.ceil(nb_validation / batch_size)) \n",
    "validation_features  =  model.predict_generator(validation_generator,  max_size_validation) \n",
    "np.save('VGG/VGG16_FE_validation_1.npy',  validation_features)\n",
    "#Test\n",
    "max_size_test = int(math.ceil(nb_test / batch_size)) \n",
    "test_features  =  model.predict_generator(test_generator,  max_size_test)\n",
    "\n",
    "#Labels Extraction\n",
    "train_labels  =  train_generator.classes  \n",
    "train_labels = to_categorical(train_labels, num_classes=nb_classes) \n",
    "validation_labels  =  validation_generator.classes  \n",
    "validation_labels  =  to_categorical(validation_labels,  num_classes=nb_classes) \n",
    "test_labels  =  test_generator.classes  \n",
    "test_labels  =  to_categorical(test_labels,  num_classes=nb_classes) \n",
    "\n",
    "#Save Features\n",
    "train_data = np.load('VGG/VGG16_FE_train_1.npy') \n",
    "validation_data  =  np.load('VGG/VGG16_FE_validation_1.npy')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Top Model Classificator\n",
    "model = Sequential() \n",
    "model.add(Flatten(input_shape=train_data.shape[1:]))\n",
    "model.add(Dense(256,  activation='relu')) \n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(nb_classes,  activation='softmax'))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Compile models\n",
    "model.compile(\n",
    "    optimizer = RMSprop(lr=2e-5),\n",
    "    loss='categorical_crossentropy',  \n",
    "    metrics=['accuracy'])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 4000 samples, validate on 1200 samples\n",
      "Epoch 1/80\n",
      "4000/4000 [==============================] - 1s 286us/step - loss: 1.3150 - acc: 0.3855 - val_loss: 1.0832 - val_acc: 0.5617\n",
      "Epoch 2/80\n",
      "4000/4000 [==============================] - 1s 156us/step - loss: 1.0629 - acc: 0.5430 - val_loss: 0.9725 - val_acc: 0.6167\n",
      "Epoch 3/80\n",
      "4000/4000 [==============================] - 1s 150us/step - loss: 0.9660 - acc: 0.5952 - val_loss: 0.9000 - val_acc: 0.6358\n",
      "Epoch 4/80\n",
      "4000/4000 [==============================] - 1s 154us/step - loss: 0.8795 - acc: 0.6530 - val_loss: 0.8499 - val_acc: 0.6492\n",
      "Epoch 5/80\n",
      "4000/4000 [==============================] - 1s 156us/step - loss: 0.8285 - acc: 0.6640 - val_loss: 0.7953 - val_acc: 0.6825\n",
      "Epoch 6/80\n",
      "4000/4000 [==============================] - 1s 136us/step - loss: 0.7826 - acc: 0.6845 - val_loss: 0.7687 - val_acc: 0.6958\n",
      "Epoch 7/80\n",
      "4000/4000 [==============================] - 1s 150us/step - loss: 0.7444 - acc: 0.7043 - val_loss: 0.7424 - val_acc: 0.7150\n",
      "Epoch 8/80\n",
      "4000/4000 [==============================] - 1s 157us/step - loss: 0.6997 - acc: 0.7298 - val_loss: 0.7183 - val_acc: 0.7375\n",
      "Epoch 9/80\n",
      "4000/4000 [==============================] - 1s 147us/step - loss: 0.6824 - acc: 0.7380 - val_loss: 0.7004 - val_acc: 0.7217\n",
      "Epoch 10/80\n",
      "4000/4000 [==============================] - 1s 151us/step - loss: 0.6375 - acc: 0.7572 - val_loss: 0.6848 - val_acc: 0.7342\n",
      "Epoch 11/80\n",
      "4000/4000 [==============================] - 1s 159us/step - loss: 0.6151 - acc: 0.7725 - val_loss: 0.6529 - val_acc: 0.7600\n",
      "Epoch 12/80\n",
      "4000/4000 [==============================] - 1s 157us/step - loss: 0.5932 - acc: 0.7802 - val_loss: 0.6676 - val_acc: 0.7333\n",
      "Epoch 13/80\n",
      "4000/4000 [==============================] - 1s 142us/step - loss: 0.5750 - acc: 0.7887 - val_loss: 0.6364 - val_acc: 0.7583\n",
      "Epoch 14/80\n",
      "4000/4000 [==============================] - 1s 146us/step - loss: 0.5520 - acc: 0.7917 - val_loss: 0.6249 - val_acc: 0.7600\n",
      "Epoch 15/80\n",
      "4000/4000 [==============================] - 1s 165us/step - loss: 0.5434 - acc: 0.8025 - val_loss: 0.6427 - val_acc: 0.7508\n",
      "Epoch 16/80\n",
      "4000/4000 [==============================] - 1s 156us/step - loss: 0.5192 - acc: 0.8190 - val_loss: 0.5979 - val_acc: 0.7767\n",
      "Epoch 17/80\n",
      "4000/4000 [==============================] - 1s 178us/step - loss: 0.5052 - acc: 0.8230 - val_loss: 0.5868 - val_acc: 0.7825\n",
      "Epoch 18/80\n",
      "4000/4000 [==============================] - 1s 185us/step - loss: 0.4860 - acc: 0.8280 - val_loss: 0.6017 - val_acc: 0.7733\n",
      "Epoch 19/80\n",
      "4000/4000 [==============================] - 1s 152us/step - loss: 0.4723 - acc: 0.8365 - val_loss: 0.6281 - val_acc: 0.7483\n",
      "Epoch 20/80\n",
      "4000/4000 [==============================] - 1s 153us/step - loss: 0.4595 - acc: 0.8430 - val_loss: 0.5945 - val_acc: 0.7692\n",
      "Epoch 21/80\n",
      "4000/4000 [==============================] - 1s 151us/step - loss: 0.4388 - acc: 0.8512 - val_loss: 0.5834 - val_acc: 0.7767\n",
      "Epoch 22/80\n",
      "4000/4000 [==============================] - 1s 148us/step - loss: 0.4295 - acc: 0.8552 - val_loss: 0.6249 - val_acc: 0.7458\n",
      "Epoch 23/80\n",
      "4000/4000 [==============================] - 1s 146us/step - loss: 0.4186 - acc: 0.8598 - val_loss: 0.5531 - val_acc: 0.7950\n",
      "Epoch 24/80\n",
      "4000/4000 [==============================] - 1s 146us/step - loss: 0.4086 - acc: 0.8653 - val_loss: 0.5525 - val_acc: 0.7892\n",
      "Epoch 25/80\n",
      "4000/4000 [==============================] - 1s 165us/step - loss: 0.3987 - acc: 0.8690 - val_loss: 0.5417 - val_acc: 0.7992\n",
      "Epoch 26/80\n",
      "4000/4000 [==============================] - 1s 159us/step - loss: 0.3906 - acc: 0.8728 - val_loss: 0.5363 - val_acc: 0.8008\n",
      "Epoch 27/80\n",
      "4000/4000 [==============================] - 1s 150us/step - loss: 0.3745 - acc: 0.8772 - val_loss: 0.5322 - val_acc: 0.8008\n",
      "Epoch 28/80\n",
      "4000/4000 [==============================] - 1s 153us/step - loss: 0.3622 - acc: 0.8865 - val_loss: 0.5341 - val_acc: 0.7958\n",
      "Epoch 29/80\n",
      "4000/4000 [==============================] - 1s 153us/step - loss: 0.3544 - acc: 0.8918 - val_loss: 0.5267 - val_acc: 0.7950\n",
      "Epoch 30/80\n",
      "4000/4000 [==============================] - 1s 161us/step - loss: 0.3426 - acc: 0.8950 - val_loss: 0.5246 - val_acc: 0.8050\n",
      "Epoch 31/80\n",
      "4000/4000 [==============================] - 1s 172us/step - loss: 0.3343 - acc: 0.8932 - val_loss: 0.5305 - val_acc: 0.7883\n",
      "Epoch 32/80\n",
      "4000/4000 [==============================] - 1s 175us/step - loss: 0.3224 - acc: 0.9035 - val_loss: 0.5214 - val_acc: 0.7975\n",
      "Epoch 33/80\n",
      "4000/4000 [==============================] - 1s 145us/step - loss: 0.3181 - acc: 0.9062 - val_loss: 0.5144 - val_acc: 0.8025\n",
      "Epoch 34/80\n",
      "4000/4000 [==============================] - 1s 144us/step - loss: 0.3140 - acc: 0.9080 - val_loss: 0.5213 - val_acc: 0.7983\n",
      "Epoch 35/80\n",
      "4000/4000 [==============================] - 1s 165us/step - loss: 0.3092 - acc: 0.9057 - val_loss: 0.5163 - val_acc: 0.8008\n",
      "Epoch 36/80\n",
      "4000/4000 [==============================] - 1s 161us/step - loss: 0.2997 - acc: 0.9110 - val_loss: 0.5464 - val_acc: 0.7842\n",
      "Epoch 37/80\n",
      "4000/4000 [==============================] - 1s 142us/step - loss: 0.2923 - acc: 0.9153 - val_loss: 0.5184 - val_acc: 0.7958\n",
      "Epoch 38/80\n",
      "4000/4000 [==============================] - 1s 152us/step - loss: 0.2814 - acc: 0.9193 - val_loss: 0.5174 - val_acc: 0.8025\n",
      "Epoch 39/80\n",
      "4000/4000 [==============================] - 1s 161us/step - loss: 0.2742 - acc: 0.9203 - val_loss: 0.5058 - val_acc: 0.8033\n",
      "Epoch 40/80\n",
      "4000/4000 [==============================] - 1s 148us/step - loss: 0.2709 - acc: 0.9237 - val_loss: 0.5033 - val_acc: 0.8058\n",
      "Epoch 41/80\n",
      "4000/4000 [==============================] - 1s 143us/step - loss: 0.2658 - acc: 0.9265 - val_loss: 0.5129 - val_acc: 0.8017\n",
      "Epoch 42/80\n",
      "4000/4000 [==============================] - 1s 144us/step - loss: 0.2545 - acc: 0.9300 - val_loss: 0.5084 - val_acc: 0.8083\n",
      "Epoch 43/80\n",
      "4000/4000 [==============================] - 1s 164us/step - loss: 0.2502 - acc: 0.9327 - val_loss: 0.5091 - val_acc: 0.8017\n",
      "Epoch 44/80\n",
      "4000/4000 [==============================] - 1s 137us/step - loss: 0.2437 - acc: 0.9390 - val_loss: 0.4951 - val_acc: 0.8083\n",
      "Epoch 45/80\n",
      "4000/4000 [==============================] - 1s 149us/step - loss: 0.2392 - acc: 0.9340 - val_loss: 0.4975 - val_acc: 0.8100\n",
      "Epoch 46/80\n",
      "4000/4000 [==============================] - 1s 149us/step - loss: 0.2315 - acc: 0.9400 - val_loss: 0.5129 - val_acc: 0.8000\n",
      "Epoch 47/80\n",
      "4000/4000 [==============================] - 1s 145us/step - loss: 0.2269 - acc: 0.9403 - val_loss: 0.4940 - val_acc: 0.8117\n",
      "Epoch 48/80\n",
      "4000/4000 [==============================] - 1s 135us/step - loss: 0.2204 - acc: 0.9427 - val_loss: 0.5546 - val_acc: 0.7883\n",
      "Epoch 49/80\n",
      "4000/4000 [==============================] - 1s 147us/step - loss: 0.2142 - acc: 0.9493 - val_loss: 0.5032 - val_acc: 0.8100\n",
      "Epoch 50/80\n",
      "4000/4000 [==============================] - 1s 157us/step - loss: 0.2157 - acc: 0.9423 - val_loss: 0.4896 - val_acc: 0.8183\n",
      "Epoch 51/80\n",
      "4000/4000 [==============================] - 1s 153us/step - loss: 0.2056 - acc: 0.9453 - val_loss: 0.5124 - val_acc: 0.8075\n",
      "Epoch 52/80\n",
      "4000/4000 [==============================] - 1s 154us/step - loss: 0.2037 - acc: 0.9502 - val_loss: 0.4941 - val_acc: 0.8117\n",
      "Epoch 53/80\n",
      "4000/4000 [==============================] - 1s 172us/step - loss: 0.2010 - acc: 0.9513 - val_loss: 0.4909 - val_acc: 0.8125\n",
      "Epoch 54/80\n",
      "4000/4000 [==============================] - 1s 170us/step - loss: 0.1922 - acc: 0.9540 - val_loss: 0.4947 - val_acc: 0.8150\n",
      "Epoch 55/80\n",
      "4000/4000 [==============================] - 1s 145us/step - loss: 0.1870 - acc: 0.9562 - val_loss: 0.5279 - val_acc: 0.7942\n",
      "Epoch 56/80\n",
      "4000/4000 [==============================] - 1s 156us/step - loss: 0.1833 - acc: 0.9535 - val_loss: 0.5021 - val_acc: 0.8058\n",
      "Epoch 57/80\n",
      "4000/4000 [==============================] - 1s 172us/step - loss: 0.1803 - acc: 0.9592 - val_loss: 0.5064 - val_acc: 0.8067\n",
      "Epoch 58/80\n",
      "4000/4000 [==============================] - 1s 153us/step - loss: 0.1746 - acc: 0.9620 - val_loss: 0.5014 - val_acc: 0.8092\n",
      "Epoch 59/80\n",
      "4000/4000 [==============================] - 1s 159us/step - loss: 0.1743 - acc: 0.9603 - val_loss: 0.4915 - val_acc: 0.8092\n",
      "Epoch 60/80\n",
      "4000/4000 [==============================] - 1s 166us/step - loss: 0.1667 - acc: 0.9635 - val_loss: 0.4919 - val_acc: 0.8058\n",
      "Epoch 61/80\n",
      "4000/4000 [==============================] - 1s 149us/step - loss: 0.1644 - acc: 0.9655 - val_loss: 0.5028 - val_acc: 0.8050\n",
      "Epoch 62/80\n",
      "4000/4000 [==============================] - 1s 135us/step - loss: 0.1596 - acc: 0.9625 - val_loss: 0.5150 - val_acc: 0.8058\n",
      "Epoch 63/80\n",
      "4000/4000 [==============================] - 1s 138us/step - loss: 0.1529 - acc: 0.9695 - val_loss: 0.4842 - val_acc: 0.8108\n",
      "Epoch 64/80\n",
      "4000/4000 [==============================] - 1s 150us/step - loss: 0.1508 - acc: 0.9655 - val_loss: 0.5141 - val_acc: 0.8033\n",
      "Epoch 65/80\n",
      "4000/4000 [==============================] - 1s 153us/step - loss: 0.1465 - acc: 0.9682 - val_loss: 0.4900 - val_acc: 0.8133\n",
      "Epoch 66/80\n",
      "4000/4000 [==============================] - 1s 154us/step - loss: 0.1463 - acc: 0.9702 - val_loss: 0.5152 - val_acc: 0.8008\n",
      "Epoch 67/80\n",
      "4000/4000 [==============================] - 1s 149us/step - loss: 0.1434 - acc: 0.9675 - val_loss: 0.4900 - val_acc: 0.8108\n",
      "Epoch 68/80\n",
      "4000/4000 [==============================] - 1s 159us/step - loss: 0.1341 - acc: 0.9765 - val_loss: 0.5014 - val_acc: 0.8083\n",
      "Epoch 69/80\n",
      "4000/4000 [==============================] - 1s 136us/step - loss: 0.1339 - acc: 0.9715 - val_loss: 0.4969 - val_acc: 0.8083\n",
      "Epoch 70/80\n",
      "4000/4000 [==============================] - 1s 140us/step - loss: 0.1329 - acc: 0.9758 - val_loss: 0.5199 - val_acc: 0.8008\n",
      "Epoch 71/80\n",
      "4000/4000 [==============================] - 1s 154us/step - loss: 0.1282 - acc: 0.9735 - val_loss: 0.4956 - val_acc: 0.8133\n",
      "Epoch 72/80\n",
      "4000/4000 [==============================] - 1s 158us/step - loss: 0.1263 - acc: 0.9790 - val_loss: 0.4918 - val_acc: 0.8133\n",
      "Epoch 73/80\n",
      "4000/4000 [==============================] - 1s 126us/step - loss: 0.1201 - acc: 0.9785 - val_loss: 0.4890 - val_acc: 0.8125\n",
      "Epoch 74/80\n",
      "4000/4000 [==============================] - 1s 131us/step - loss: 0.1195 - acc: 0.9795 - val_loss: 0.5189 - val_acc: 0.8058\n",
      "Epoch 75/80\n",
      "4000/4000 [==============================] - 1s 156us/step - loss: 0.1191 - acc: 0.9785 - val_loss: 0.5078 - val_acc: 0.8008\n",
      "Epoch 76/80\n",
      "4000/4000 [==============================] - 1s 153us/step - loss: 0.1118 - acc: 0.9780 - val_loss: 0.4970 - val_acc: 0.8150\n",
      "Epoch 77/80\n",
      "4000/4000 [==============================] - 1s 150us/step - loss: 0.1100 - acc: 0.9808 - val_loss: 0.5017 - val_acc: 0.8083\n",
      "Epoch 78/80\n",
      "4000/4000 [==============================] - 1s 164us/step - loss: 0.1055 - acc: 0.9852 - val_loss: 0.4927 - val_acc: 0.8117\n",
      "Epoch 79/80\n",
      "4000/4000 [==============================] - 1s 163us/step - loss: 0.1095 - acc: 0.9790 - val_loss: 0.5027 - val_acc: 0.8108\n",
      "Epoch 80/80\n",
      "4000/4000 [==============================] - 1s 136us/step - loss: 0.1015 - acc: 0.9830 - val_loss: 0.5033 - val_acc: 0.8100\n"
     ]
    }
   ],
   "source": [
    "#Train model\n",
    "history = model.fit(\n",
    "    train_data, \n",
    "    train_labels,\n",
    "    epochs=epochs,  \n",
    "    batch_size=batch_size, \n",
    "    validation_data=(validation_data, validation_labels))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# h5py \n",
    "model.save_weights('VGG/VGG16_FE_model_1.h5')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def print_graphics (history):\n",
    "    #Mostramos otro tipo de grafico\n",
    "    acc = history.history['acc']\n",
    "    val_acc = history.history['val_acc']\n",
    "    loss = history.history['loss']\n",
    "    val_loss = history.history['val_loss']\n",
    "\n",
    "    epochs = range(1, len(acc)+ 1)\n",
    "\n",
    "    plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "    plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "    #plt.tittle('Trainning and validation accuracy')\n",
    "    plt.legend()\n",
    "\n",
    "    plt.figure()\n",
    "    plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "    plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "    #plt.tittle('Trainning and validation loss')\n",
    "    plt.legend()\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def print_evaluate_model(model, history):\n",
    "    #Evaluate model\n",
    "    (loss, acc) = model.evaluate(\n",
    "        test_features, test_labels, \n",
    "        batch_size=batch_size, \n",
    "        verbose=0)\n",
    "\n",
    "    print(\"acc: {0:.2f}% - loss: {1:f}\".format(acc * 100, loss))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "acc: 79.17% - loss: 0.565496\n"
     ]
    }
   ],
   "source": [
    "print_evaluate_model(model, history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff54829d0b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff548332fd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print_graphics (history)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}