Mobilenet.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np  \n",
    "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 import applications  \n",
    "from keras.utils.np_utils import to_categorical  \n",
    "import matplotlib.pyplot as plt  \n",
    "import math  \n",
    "\n",
    "# dimensions of our images \n",
    "img_width, img_height = 224, 224  \n",
    "\n",
    "weights_path = 'mobilenet_model.h5'  \n",
    "train_data_dir = 'train_val_test/train'  \n",
    "validation_data_dir = 'train_val_test/val'  \n",
    "test_data_dir = 'train_val_test/test'  \n",
    "   \n",
    " # number of epochs to train top model  \n",
    "epochs = 50  \n",
    " # batch size used by flow_from_directory and predict_generator  \n",
    "batch_size = 8  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\juigu\\Anaconda2\\envs\\Python3\\lib\\site-packages\\keras_applications\\mobilenet.py:207: UserWarning: MobileNet shape is undefined. Weights for input shape (224, 224) will be loaded.\n",
      "  warnings.warn('MobileNet shape is undefined.'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 8185 images belonging to 120 classes.\n"
     ]
    }
   ],
   "source": [
    "model = applications.mobilenet.MobileNet(include_top=False, weights='imagenet')\n",
    "datagen = ImageDataGenerator(rescale=1. / 255)  \n",
    "   \n",
    "generator = datagen.flow_from_directory(  \n",
    "     train_data_dir,  \n",
    "     target_size=(img_width, img_height),  \n",
    "     batch_size=batch_size,  \n",
    "     class_mode=None,  \n",
    "     shuffle=False)  \n",
    "   \n",
    "nb_train_samples = len(generator.filenames)  \n",
    "num_classes = len(generator.class_indices)  \n",
    "   \n",
    "predict_size_train = int(math.ceil(nb_train_samples / batch_size))  \n",
    "   \n",
    "bottleneck_features_train = model.predict_generator(  \n",
    "     generator, predict_size_train)  \n",
    "   \n",
    "np.save('bottleneck_features_train_mob.npy', bottleneck_features_train)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 1015 images belonging to 120 classes.\n"
     ]
    }
   ],
   "source": [
    "generator = datagen.flow_from_directory(  \n",
    "     validation_data_dir,  \n",
    "     target_size=(img_width, img_height),  \n",
    "     batch_size=batch_size,  \n",
    "     class_mode=None,  \n",
    "     shuffle=False)  \n",
    "   \n",
    "nb_validation_samples = len(generator.filenames)  \n",
    "   \n",
    "predict_size_validation = int(math.ceil(nb_validation_samples / batch_size))  \n",
    "   \n",
    "bottleneck_features_validation = model.predict_generator(  \n",
    "     generator, predict_size_validation)  \n",
    "   \n",
    "np.save('bottleneck_features_validation_mob.npy', bottleneck_features_validation) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 8185 images belonging to 120 classes.\n"
     ]
    }
   ],
   "source": [
    "datagen_top = ImageDataGenerator(rescale=1./255)  \n",
    "generator_top = datagen_top.flow_from_directory(  \n",
    "         train_data_dir,  \n",
    "         target_size=(img_width, img_height),  \n",
    "         batch_size=batch_size,  \n",
    "         class_mode='categorical',  \n",
    "         shuffle=False)  \n",
    "   \n",
    "nb_train_samples = len(generator_top.filenames)  \n",
    "num_classes = len(generator_top.class_indices)  \n",
    "   \n",
    " # load the bottleneck features saved earlier  \n",
    "train_data = np.load('bottleneck_features_train_mob.npy')  \n",
    "   \n",
    " # get the class lebels for the training data, in the original order  \n",
    "train_labels = generator_top.classes  \n",
    "   \n",
    "# convert the training labels to categorical vectors  \n",
    "train_labels = to_categorical(train_labels, num_classes=num_classes)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 1015 images belonging to 120 classes.\n"
     ]
    }
   ],
   "source": [
    "generator_top = datagen_top.flow_from_directory(  \n",
    "         validation_data_dir,  \n",
    "         target_size=(img_width, img_height),  \n",
    "         batch_size=batch_size,  \n",
    "         class_mode=None,  \n",
    "         shuffle=False)  \n",
    "   \n",
    "nb_validation_samples = len(generator_top.filenames)  \n",
    "   \n",
    "validation_data = np.load('bottleneck_features_validation_mob.npy')  \n",
    "   \n",
    "validation_labels = generator_top.classes  \n",
    "validation_labels = to_categorical(validation_labels, num_classes=num_classes) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 8185 samples, validate on 1015 samples\n",
      "Epoch 1/50\n",
      "8185/8185 [==============================] - 319s 39ms/step - loss: 15.9167 - acc: 0.0105 - val_loss: 15.9911 - val_acc: 0.0079\n",
      "Epoch 2/50\n",
      "8185/8185 [==============================] - 207s 25ms/step - loss: 15.9190 - acc: 0.0106 - val_loss: 15.7230 - val_acc: 0.0187\n",
      "Epoch 3/50\n",
      "8185/8185 [==============================] - 189s 23ms/step - loss: 15.8003 - acc: 0.0177 - val_loss: 15.6320 - val_acc: 0.0296\n",
      "Epoch 4/50\n",
      "8185/8185 [==============================] - 215s 26ms/step - loss: 15.7088 - acc: 0.0225 - val_loss: 15.6266 - val_acc: 0.0286\n",
      "Epoch 5/50\n",
      "8185/8185 [==============================] - 219s 27ms/step - loss: 15.6778 - acc: 0.0254 - val_loss: 15.5459 - val_acc: 0.0315\n",
      "Epoch 6/50\n",
      "8185/8185 [==============================] - 215s 26ms/step - loss: 15.5668 - acc: 0.0321 - val_loss: 15.3410 - val_acc: 0.0443\n",
      "Epoch 7/50\n",
      "8185/8185 [==============================] - 215s 26ms/step - loss: 15.4682 - acc: 0.0379 - val_loss: 15.2694 - val_acc: 0.0473\n",
      "Epoch 8/50\n",
      "8185/8185 [==============================] - 209s 26ms/step - loss: 15.3929 - acc: 0.0428 - val_loss: 15.2161 - val_acc: 0.0522\n",
      "Epoch 9/50\n",
      "8185/8185 [==============================] - 213s 26ms/step - loss: 15.2832 - acc: 0.0491 - val_loss: 15.0212 - val_acc: 0.0670\n",
      "Epoch 10/50\n",
      "8185/8185 [==============================] - 210s 26ms/step - loss: 15.2611 - acc: 0.0508 - val_loss: 15.0275 - val_acc: 0.0670\n",
      "Epoch 11/50\n",
      "8185/8185 [==============================] - 216s 26ms/step - loss: 15.1040 - acc: 0.0601 - val_loss: 14.8634 - val_acc: 0.0719\n",
      "Epoch 12/50\n",
      "8185/8185 [==============================] - 222s 27ms/step - loss: 14.9707 - acc: 0.0685 - val_loss: 14.5583 - val_acc: 0.0926\n",
      "Epoch 13/50\n",
      "8185/8185 [==============================] - 220s 27ms/step - loss: 14.7984 - acc: 0.0786 - val_loss: 14.5051 - val_acc: 0.0975\n",
      "Epoch 14/50\n",
      "8185/8185 [==============================] - 223s 27ms/step - loss: 14.7816 - acc: 0.0786 - val_loss: 14.3988 - val_acc: 0.1064\n",
      "Epoch 15/50\n",
      "8185/8185 [==============================] - 227s 28ms/step - loss: 14.6515 - acc: 0.0869 - val_loss: 14.3077 - val_acc: 0.1103\n",
      "Epoch 16/50\n",
      "8185/8185 [==============================] - 221s 27ms/step - loss: 14.5582 - acc: 0.0932 - val_loss: 14.4249 - val_acc: 0.1015\n",
      "Epoch 17/50\n",
      "8185/8185 [==============================] - 228s 28ms/step - loss: 14.4355 - acc: 0.0995 - val_loss: 14.1749 - val_acc: 0.1172\n",
      "Epoch 18/50\n",
      "8185/8185 [==============================] - 227s 28ms/step - loss: 14.3341 - acc: 0.1062 - val_loss: 14.1272 - val_acc: 0.1192\n",
      "Epoch 19/50\n",
      "8185/8185 [==============================] - 222s 27ms/step - loss: 14.3036 - acc: 0.1085 - val_loss: 14.0945 - val_acc: 0.1222\n",
      "Epoch 20/50\n",
      "8185/8185 [==============================] - 227s 28ms/step - loss: 14.2360 - acc: 0.1136 - val_loss: 13.9651 - val_acc: 0.1310\n",
      "Epoch 21/50\n",
      "8185/8185 [==============================] - 216s 26ms/step - loss: 14.1129 - acc: 0.1202 - val_loss: 13.8060 - val_acc: 0.1399\n",
      "Epoch 22/50\n",
      "8185/8185 [==============================] - 212s 26ms/step - loss: 13.9814 - acc: 0.1277 - val_loss: 13.7618 - val_acc: 0.1399\n",
      "Epoch 23/50\n",
      "8185/8185 [==============================] - 214s 26ms/step - loss: 13.9957 - acc: 0.1266 - val_loss: 13.7115 - val_acc: 0.1458\n",
      "Epoch 24/50\n",
      "8185/8185 [==============================] - 209s 25ms/step - loss: 13.8444 - acc: 0.1368 - val_loss: 13.4423 - val_acc: 0.1606\n",
      "Epoch 25/50\n",
      "8185/8185 [==============================] - 213s 26ms/step - loss: 13.7382 - acc: 0.1410 - val_loss: 13.5246 - val_acc: 0.1576\n",
      "Epoch 26/50\n",
      "8185/8185 [==============================] - 213s 26ms/step - loss: 13.6437 - acc: 0.1475 - val_loss: 13.4218 - val_acc: 0.1635\n",
      "Epoch 27/50\n",
      "8185/8185 [==============================] - 208s 25ms/step - loss: 13.4860 - acc: 0.1570 - val_loss: 13.3188 - val_acc: 0.1685\n",
      "Epoch 28/50\n",
      "8185/8185 [==============================] - 213s 26ms/step - loss: 13.4965 - acc: 0.1571 - val_loss: 13.0995 - val_acc: 0.1793\n",
      "Epoch 29/50\n",
      "8185/8185 [==============================] - 212s 26ms/step - loss: 13.3918 - acc: 0.1607 - val_loss: 12.9311 - val_acc: 0.1882\n",
      "Epoch 30/50\n",
      "8185/8185 [==============================] - 209s 26ms/step - loss: 13.3399 - acc: 0.1659 - val_loss: 12.8011 - val_acc: 0.1970\n",
      "Epoch 31/50\n",
      "8185/8185 [==============================] - 215s 26ms/step - loss: 13.2180 - acc: 0.1745 - val_loss: 12.5816 - val_acc: 0.2148\n",
      "Epoch 32/50\n",
      "8185/8185 [==============================] - 211s 26ms/step - loss: 13.0773 - acc: 0.1812 - val_loss: 12.6607 - val_acc: 0.2089\n",
      "Epoch 33/50\n",
      "8185/8185 [==============================] - 210s 26ms/step - loss: 12.9387 - acc: 0.1891 - val_loss: 12.6494 - val_acc: 0.2069\n",
      "Epoch 34/50\n",
      "8185/8185 [==============================] - 214s 26ms/step - loss: 12.8336 - acc: 0.1980 - val_loss: 12.5809 - val_acc: 0.2148\n",
      "Epoch 35/50\n",
      "8185/8185 [==============================] - 211s 26ms/step - loss: 12.6733 - acc: 0.2072 - val_loss: 12.3904 - val_acc: 0.2266\n",
      "Epoch 36/50\n",
      "8185/8185 [==============================] - 211s 26ms/step - loss: 12.5919 - acc: 0.2128 - val_loss: 12.4255 - val_acc: 0.2236\n",
      "Epoch 37/50\n",
      "8185/8185 [==============================] - 214s 26ms/step - loss: 12.5350 - acc: 0.2154 - val_loss: 12.2514 - val_acc: 0.2325\n",
      "Epoch 38/50\n",
      "8185/8185 [==============================] - 210s 26ms/step - loss: 12.4812 - acc: 0.2182 - val_loss: 12.3102 - val_acc: 0.2296\n",
      "Epoch 39/50\n",
      "8185/8185 [==============================] - 212s 26ms/step - loss: 12.5100 - acc: 0.2171 - val_loss: 12.1518 - val_acc: 0.2394\n",
      "Epoch 40/50\n",
      "8185/8185 [==============================] - 216s 26ms/step - loss: 12.4072 - acc: 0.2248 - val_loss: 12.1786 - val_acc: 0.2394\n",
      "Epoch 41/50\n",
      "8185/8185 [==============================] - 209s 26ms/step - loss: 12.2612 - acc: 0.2337 - val_loss: 12.0920 - val_acc: 0.2424\n",
      "Epoch 42/50\n",
      "8185/8185 [==============================] - 213s 26ms/step - loss: 12.1874 - acc: 0.2367 - val_loss: 12.0781 - val_acc: 0.2443\n",
      "Epoch 43/50\n",
      "8185/8185 [==============================] - 215s 26ms/step - loss: 12.1906 - acc: 0.2375 - val_loss: 12.0313 - val_acc: 0.2453\n",
      "Epoch 44/50\n",
      "8185/8185 [==============================] - 209s 26ms/step - loss: 12.1024 - acc: 0.2429 - val_loss: 12.1171 - val_acc: 0.2394\n",
      "Epoch 45/50\n",
      "8185/8185 [==============================] - 214s 26ms/step - loss: 12.0952 - acc: 0.2425 - val_loss: 12.0297 - val_acc: 0.2453\n",
      "Epoch 46/50\n",
      "8185/8185 [==============================] - 213s 26ms/step - loss: 12.0667 - acc: 0.2456 - val_loss: 12.0701 - val_acc: 0.2433\n",
      "Epoch 47/50\n",
      "8185/8185 [==============================] - 210s 26ms/step - loss: 11.9694 - acc: 0.2512 - val_loss: 11.9353 - val_acc: 0.2522\n",
      "Epoch 48/50\n",
      "8185/8185 [==============================] - 214s 26ms/step - loss: 11.9964 - acc: 0.2498 - val_loss: 11.7812 - val_acc: 0.2611\n",
      "Epoch 49/50\n",
      "8185/8185 [==============================] - 211s 26ms/step - loss: 11.9566 - acc: 0.2531 - val_loss: 11.9152 - val_acc: 0.2562\n",
      "Epoch 50/50\n",
      "8185/8185 [==============================] - 210s 26ms/step - loss: 11.8782 - acc: 0.2568 - val_loss: 12.1029 - val_acc: 0.2433\n",
      "1015/1015 [==============================] - 4s 4ms/step\n",
      "[INFO] accuracy: 24.33%\n",
      "[INFO] Loss: 12.102912517310969\n"
     ]
    }
   ],
   "source": [
    "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(num_classes, activation='softmax'))  \n",
    "   \n",
    "model.compile(optimizer='rmsprop',  \n",
    "              loss='categorical_crossentropy', metrics=['accuracy'])  \n",
    "   \n",
    "history = model.fit(train_data, train_labels,  \n",
    "          epochs=epochs,  \n",
    "          batch_size=batch_size,  \n",
    "          validation_data=(validation_data, validation_labels))  \n",
    "   \n",
    "model.save_weights(weights_path)  \n",
    "   \n",
    "(eval_loss, eval_accuracy) = model.evaluate(  \n",
    "     validation_data, validation_labels, batch_size=batch_size, verbose=1)\n",
    "\n",
    "print(\"[INFO] accuracy: {:.2f}%\".format(eval_accuracy * 100))  \n",
    "print(\"[INFO] Loss: {}\".format(eval_loss)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for accuracy  \n",
    "   \n",
    "plt.subplot(211)  \n",
    "plt.plot(history.history['acc'])  \n",
    "plt.plot(history.history['val_acc'])  \n",
    "plt.title('MobileNet model accuracy')  \n",
    "plt.ylabel('accuracy')  \n",
    "plt.xlabel('epoch')  \n",
    "plt.legend(['train', 'test'], loc='upper left')  \n",
    "  \n",
    "# summarize history for loss  \n",
    "   \n",
    "plt.subplot(212)  \n",
    "plt.plot(history.history['loss'])  \n",
    "plt.plot(history.history['val_loss'])  \n",
    "plt.title('model loss')  \n",
    "plt.ylabel('loss')  \n",
    "plt.xlabel('epoch')  \n",
    "plt.legend(['train', 'test'], loc='upper left')  \n",
    "plt.show() "
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}