generator_comparison.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## This notebook is made for comparison between dataset provided by loading the whole one into notebook vs the using a generator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[name: \"/device:CPU:0\"\n",
       " device_type: \"CPU\"\n",
       " memory_limit: 268435456\n",
       " locality {\n",
       " }\n",
       " incarnation: 11482694941060484880, name: \"/device:GPU:0\"\n",
       " device_type: \"GPU\"\n",
       " memory_limit: 356122624\n",
       " locality {\n",
       "   bus_id: 1\n",
       " }\n",
       " incarnation: 16073676007118189501\n",
       " physical_device_desc: \"device: 0, name: Tesla K80, pci bus id: 0000:00:1e.0, compute capability: 3.7\"]"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.python.client import device_lib\n",
    "sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))\n",
    "device_lib.list_local_devices()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras.layers import Activation, Dropout, Flatten, Dense\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "from keras.utils.vis_utils import plot_model\n",
    "from sklearn.utils import shuffle\n",
    "from keras import backend as K\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "width, height = 64, 64\n",
    "train_data_dir = '../zdjecia_gen/train'\n",
    "test_data_dir = '../zdjecia_gen/test'\n",
    "nb_train_samples = 40320\n",
    "nb_test_samples = 10080\n",
    "epochs = 20\n",
    "batch_size = 64\n",
    "\n",
    "input_shape = (width, height, 3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/keras/backend/tensorflow_backend.py:1264: calling reduce_prod (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Conv2D(32, (5, 5), input_shape=input_shape))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Conv2D(32, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Conv2D(64, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(1024))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(6, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/keras/backend/tensorflow_backend.py:2885: calling reduce_sum (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/keras/backend/tensorflow_backend.py:1349: calling reduce_mean (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 60, 60, 32)        2432      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 60, 60, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 30, 30, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 28, 28, 32)        9248      \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 28, 28, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 14, 14, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 12, 12, 64)        18496     \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 12, 12, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 2304)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1024)              2360320   \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 64)                65600     \n",
      "_________________________________________________________________\n",
      "activation_5 (Activation)    (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 6)                 390       \n",
      "=================================================================\n",
      "Total params: 2,456,486\n",
      "Trainable params: 2,456,486\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "# summarize model and apply checkpoints\n",
    "print(model.summary())\n",
    "filepath = 'generator_comparison/comparison-{epoch:02d}-{loss:.4f}.h5'\n",
    "checkpoint = ModelCheckpoint(filepath, monitor='loss', verbose=1, save_best_only=True, mode='min')\n",
    "callbacks_list = [checkpoint]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_datagen = ImageDataGenerator(\n",
    "    rescale=1. / 255)\n",
    "test_datagen = ImageDataGenerator(rescale=1. / 255)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 40320 images belonging to 6 classes.\n",
      "Found 10080 images belonging to 6 classes.\n"
     ]
    }
   ],
   "source": [
    "train_generator = train_datagen.flow_from_directory(\n",
    "    train_data_dir,\n",
    "    target_size=(width, height),\n",
    "    color_mode='rgb',\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical',\n",
    "    classes=['1', '2', '3', '4', '5', '6'],\n",
    "    shuffle=True,\n",
    "    seed=2018)\n",
    "\n",
    "test_generator = test_datagen.flow_from_directory(\n",
    "    test_data_dir,\n",
    "    target_size=(width, height),\n",
    "    color_mode='rgb',\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical',\n",
    "    classes=['1', '2', '3', '4', '5', '6'],\n",
    "    shuffle=True,\n",
    "    seed=2018)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "629/630 [============================>.] - ETA: 0s - loss: 1.2184 - acc: 0.4623Epoch 00001: loss improved from inf to 1.21773, saving model to generator_comparison/comparison-01-1.2177.h5\n",
      "630/630 [==============================] - 19s 30ms/step - loss: 1.2177 - acc: 0.4625 - val_loss: 0.5454 - val_acc: 0.7941\n",
      "Epoch 2/20\n",
      "627/630 [============================>.] - ETA: 0s - loss: 0.4016 - acc: 0.8413Epoch 00002: loss improved from 1.21773 to 0.40077, saving model to generator_comparison/comparison-02-0.4008.h5\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.4008 - acc: 0.8417 - val_loss: 0.2348 - val_acc: 0.9092\n",
      "Epoch 3/20\n",
      "627/630 [============================>.] - ETA: 0s - loss: 0.2031 - acc: 0.9200Epoch 00003: loss improved from 0.40077 to 0.20335, saving model to generator_comparison/comparison-03-0.2033.h5\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.2033 - acc: 0.9198 - val_loss: 0.1231 - val_acc: 0.9517\n",
      "Epoch 4/20\n",
      "629/630 [============================>.] - ETA: 0s - loss: 0.1297 - acc: 0.9507Epoch 00004: loss improved from 0.20335 to 0.12965, saving model to generator_comparison/comparison-04-0.1297.h5\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.1297 - acc: 0.9508 - val_loss: 0.0886 - val_acc: 0.9661\n",
      "Epoch 5/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0903 - acc: 0.9661Epoch 00005: loss improved from 0.12965 to 0.09025, saving model to generator_comparison/comparison-05-0.0902.h5\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.0902 - acc: 0.9661 - val_loss: 0.0753 - val_acc: 0.9710\n",
      "Epoch 6/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0643 - acc: 0.9771Epoch 00006: loss improved from 0.09025 to 0.06438, saving model to generator_comparison/comparison-06-0.0644.h5\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0644 - acc: 0.9771 - val_loss: 0.0543 - val_acc: 0.9796\n",
      "Epoch 7/20\n",
      "629/630 [============================>.] - ETA: 0s - loss: 0.0576 - acc: 0.9799Epoch 00007: loss improved from 0.06438 to 0.05770, saving model to generator_comparison/comparison-07-0.0577.h5\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.0577 - acc: 0.9799 - val_loss: 0.0770 - val_acc: 0.9701\n",
      "Epoch 8/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0444 - acc: 0.9848Epoch 00008: loss improved from 0.05770 to 0.04427, saving model to generator_comparison/comparison-08-0.0443.h5\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0443 - acc: 0.9849 - val_loss: 0.0541 - val_acc: 0.9813\n",
      "Epoch 9/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0365 - acc: 0.9877Epoch 00009: loss improved from 0.04427 to 0.03659, saving model to generator_comparison/comparison-09-0.0366.h5\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.0366 - acc: 0.9876 - val_loss: 0.0691 - val_acc: 0.9758\n",
      "Epoch 10/20\n",
      "629/630 [============================>.] - ETA: 0s - loss: 0.0407 - acc: 0.9863Epoch 00010: loss did not improve\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0406 - acc: 0.9864 - val_loss: 0.0439 - val_acc: 0.9853\n",
      "Epoch 11/20\n",
      "627/630 [============================>.] - ETA: 0s - loss: 0.0315 - acc: 0.9903Epoch 00011: loss improved from 0.03659 to 0.03138, saving model to generator_comparison/comparison-11-0.0314.h5\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.0314 - acc: 0.9903 - val_loss: 0.0327 - val_acc: 0.9882\n",
      "Epoch 12/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0231 - acc: 0.9926Epoch 00012: loss improved from 0.03138 to 0.02319, saving model to generator_comparison/comparison-12-0.0232.h5\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.0232 - acc: 0.9926 - val_loss: 0.0461 - val_acc: 0.9845\n",
      "Epoch 13/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0346 - acc: 0.9894Epoch 00013: loss did not improve\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.0345 - acc: 0.9894 - val_loss: 0.0541 - val_acc: 0.9823\n",
      "Epoch 14/20\n",
      "627/630 [============================>.] - ETA: 0s - loss: 0.0239 - acc: 0.9925Epoch 00014: loss did not improve\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0238 - acc: 0.9926 - val_loss: 0.0263 - val_acc: 0.9913\n",
      "Epoch 15/20\n",
      "627/630 [============================>.] - ETA: 0s - loss: 0.0171 - acc: 0.9940Epoch 00015: loss improved from 0.02319 to 0.01704, saving model to generator_comparison/comparison-15-0.0170.h5\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0170 - acc: 0.9940 - val_loss: 0.0596 - val_acc: 0.9845\n",
      "Epoch 16/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0302 - acc: 0.9908Epoch 00016: loss did not improve\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0302 - acc: 0.9908 - val_loss: 0.0423 - val_acc: 0.9879\n",
      "Epoch 17/20\n",
      "629/630 [============================>.] - ETA: 0s - loss: 0.0170 - acc: 0.9950Epoch 00017: loss improved from 0.01704 to 0.01693, saving model to generator_comparison/comparison-17-0.0169.h5\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0169 - acc: 0.9950 - val_loss: 0.0515 - val_acc: 0.9859\n",
      "Epoch 18/20\n",
      "628/630 [============================>.] - ETA: 0s - loss: 0.0212 - acc: 0.9936Epoch 00018: loss did not improve\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0211 - acc: 0.9936 - val_loss: 0.0308 - val_acc: 0.9907\n",
      "Epoch 19/20\n",
      "627/630 [============================>.] - ETA: 0s - loss: 0.0165 - acc: 0.9947Epoch 00019: loss improved from 0.01693 to 0.01655, saving model to generator_comparison/comparison-19-0.0166.h5\n",
      "630/630 [==============================] - 18s 29ms/step - loss: 0.0166 - acc: 0.9947 - val_loss: 0.0318 - val_acc: 0.9907\n",
      "Epoch 20/20\n",
      "629/630 [============================>.] - ETA: 0s - loss: 0.0216 - acc: 0.9937Epoch 00020: loss did not improve\n",
      "630/630 [==============================] - 18s 28ms/step - loss: 0.0216 - acc: 0.9937 - val_loss: 0.0569 - val_acc: 0.9848\n"
     ]
    }
   ],
   "source": [
    "history = model.fit_generator(\n",
    "    train_generator,\n",
    "    steps_per_epoch=nb_train_samples // batch_size,\n",
    "    epochs=epochs,\n",
    "    verbose=1,\n",
    "    callbacks=callbacks_list,\n",
    "    shuffle=True,\n",
    "    validation_data=test_generator,\n",
    "    validation_steps=nb_test_samples // batch_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 98.47%\n",
      "Test loss: 0.05701799388291889\n",
      "Test accuracy 0.9847222222222223\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate_generator(test_generator)\n",
    "print(\"Accuracy: %.2f%%\" % (score[1]*100))\n",
    "print('Test loss:', score[0])\n",
    "print('Test accuracy', score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saved model weights\n"
     ]
    }
   ],
   "source": [
    "model_json = model.to_json()\n",
    "with open('comparison_model.json', 'w') as json_file:\n",
    "    json_file.write(model_json)\n",
    "    \n",
    "model.save_weights('comparison_weights.h5')\n",
    "print('Saved model weights')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff484a1a208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#summarize for accuracy\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff484a68470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#summarize for loss\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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}