Essaie.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Conv2D, MaxPooling2D, Conv2DTranspose\n",
    "from tensorflow.keras.layers import Activation, Dropout, Flatten, Dense\n",
    "from tensorflow.keras.layers import concatenate\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from tensorflow.keras import backend as K\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras.applications import VGG16\n",
    "\n",
    "\n",
    "# dimensions of our images.\n",
    "img_width, img_height = 150, 150\n",
    "\n",
    "train_data_dir = 'C:/Users/Alou/Desktop/test/dataset/training_set'\n",
    "validation_data_dir = 'C:/Users/Alou/Desktop/test/dataset/test_set'\n",
    "nb_train_samples = 8000\n",
    "nb_validation_samples = 2000\n",
    "epochs = 10\n",
    "batch_size = 16\n",
    "\n",
    "if K.image_data_format() == 'channels_first':\n",
    "    input_shape = (3, img_width, img_height)\n",
    "else:\n",
    "    input_shape = (img_width, img_height, 3)\n",
    "    \n",
    "\n",
    "\n",
    "def UEfficientNet(input_shape=(None, None, 3),dropout_rate=0.1):\n",
    "\n",
    "    backbone = VGG16(weights='imagenet',\n",
    "                            include_top=False,\n",
    "                            input_shape=input_shape)\n",
    "    input_layer = backbone.input\n",
    "    start_neurons = 16\n",
    "\n",
    "    conv4 = backbone.layers[342].output\n",
    "    conv4 = LeakyReLU(alpha=0.1)(conv4)\n",
    "    pool4 = MaxPooling2D((2, 2))(conv4)\n",
    "    pool4 = Dropout(dropout_rate)(pool4)\n",
    "    \n",
    "     # Middle\n",
    "    convm = Conv2D(start_neurons * 32, (3, 3), activation=None, padding=\"same\")(pool4)\n",
    "    convm = residual_block(convm,start_neurons * 32)\n",
    "    convm = residual_block(convm,start_neurons * 32)\n",
    "    convm = LeakyReLU(alpha=0.1)(convm)\n",
    "    \n",
    "    deconv4 = Conv2DTranspose(start_neurons * 16, (3, 3), strides=(2, 2), padding=\"same\")(convm)\n",
    "    uconv4 = concatenate([deconv4, conv4])\n",
    "    uconv4 = Dropout(dropout_rate)(uconv4)\n",
    "    \n",
    "    uconv4 = Conv2D(start_neurons * 16, (3, 3), activation=None, padding=\"same\")(uconv4)\n",
    "    uconv4 = residual_block(uconv4,start_neurons * 16)\n",
    "    uconv4 = residual_block(uconv4,start_neurons * 16)\n",
    "    uconv4 = LeakyReLU(alpha=0.1)(uconv4)\n",
    "    \n",
    "    deconv3 = Conv2DTranspose(start_neurons * 8, (3, 3), strides=(2, 2), padding=\"same\")(uconv4)\n",
    "    conv3 = backbone.layers[154].output\n",
    "    uconv3 = concatenate([deconv3, conv3])    \n",
    "    uconv3 = Dropout(dropout_rate)(uconv3)\n",
    "    \n",
    "    uconv3 = Conv2D(start_neurons * 8, (3, 3), activation=None, padding=\"same\")(uconv3)\n",
    "    uconv3 = residual_block(uconv3,start_neurons * 8)\n",
    "    uconv3 = residual_block(uconv3,start_neurons * 8)\n",
    "    uconv3 = LeakyReLU(alpha=0.1)(uconv3)\n",
    "\n",
    "    deconv2 = Conv2DTranspose(start_neurons * 4, (3, 3), strides=(2, 2), padding=\"same\")(uconv3)\n",
    "    conv2 = backbone.layers[92].output\n",
    "    uconv2 = concatenate([deconv2, conv2])\n",
    "        \n",
    "    uconv2 = Dropout(0.1)(uconv2)\n",
    "    uconv2 = Conv2D(start_neurons * 4, (3, 3), activation=None, padding=\"same\")(uconv2)\n",
    "    uconv2 = residual_block(uconv2,start_neurons * 4)\n",
    "    uconv2 = residual_block(uconv2,start_neurons * 4)\n",
    "    uconv2 = LeakyReLU(alpha=0.1)(uconv2)\n",
    "    \n",
    "    deconv1 = Conv2DTranspose(start_neurons * 2, (3, 3), strides=(2, 2), padding=\"same\")(uconv2)\n",
    "    conv1 = backbone.layers[30].output\n",
    "    uconv1 = concatenate([deconv1, conv1])\n",
    "    \n",
    "    uconv1 = Dropout(0.1)(uconv1)\n",
    "    uconv1 = Conv2D(start_neurons * 2, (3, 3), activation=None, padding=\"same\")(uconv1)\n",
    "    uconv1 = residual_block(uconv1,start_neurons * 2)\n",
    "    uconv1 = residual_block(uconv1,start_neurons * 2)\n",
    "    uconv1 = LeakyReLU(alpha=0.1)(uconv1)\n",
    "    \n",
    "    uconv0 = Conv2DTranspose(start_neurons * 1, (3, 3), strides=(2, 2), padding=\"same\")(uconv1)   \n",
    "    uconv0 = Dropout(0.1)(uconv0)\n",
    "    uconv0 = Conv2D(start_neurons * 1, (3, 3), activation=None, padding=\"same\")(uconv0)\n",
    "    uconv0 = residual_block(uconv0,start_neurons * 1)\n",
    "    uconv0 = residual_block(uconv0,start_neurons * 1)\n",
    "    uconv0 = LeakyReLU(alpha=0.1)(uconv0)\n",
    "    \n",
    "    uconv0 = Dropout(dropout_rate/2)(uconv0)\n",
    "    output_layer = Conv2D(1, (1,1), padding=\"same\", activation=\"sigmoid\")(uconv0)    \n",
    "    \n",
    "    model = Model(input_layer, output_layer)\n",
    "    model.name = 'u-xception'\n",
    "\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "IndexError",
     "evalue": "list index out of range",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mIndexError\u001b[0m                                Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-12-b75f256318d3>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[0mK\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mclear_session\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[0mmodel\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mUEfficientNet\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0minput_shape\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m256\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m256\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m3\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mdropout_rate\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0.25\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;32m<ipython-input-11-6a91e1eabe8f>\u001b[0m in \u001b[0;36mUEfficientNet\u001b[1;34m(input_shape, dropout_rate)\u001b[0m\n\u001b[0;32m     35\u001b[0m     \u001b[0mstart_neurons\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;36m16\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     36\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 37\u001b[1;33m     \u001b[0mconv4\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mbackbone\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;36m342\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0moutput\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     38\u001b[0m     \u001b[0mconv4\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mLeakyReLU\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0malpha\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0.1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mconv4\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     39\u001b[0m     \u001b[0mpool4\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mMaxPooling2D\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m2\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;36m2\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mconv4\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mIndexError\u001b[0m: list index out of range"
     ]
    }
   ],
   "source": [
    "\n",
    "K.clear_session()\n",
    "model = UEfficientNet(input_shape=(256,256,3),dropout_rate=0.25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss='binary_crossentropy',\n",
    "              optimizer='adam',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# this is the augmentation configuration we will use for training\n",
    "train_datagen = ImageDataGenerator(\n",
    "    rescale=1. / 255,\n",
    "    shear_range=0.2,\n",
    "    zoom_range=0.2,\n",
    "    horizontal_flip=True)\n",
    "\n",
    "# this is the augmentation configuration we will use for testing:\n",
    "# only rescaling\n",
    "test_datagen = ImageDataGenerator(rescale=1. / 255)\n",
    "\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "    train_data_dir,\n",
    "    target_size=(img_width, img_height),\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical')\n",
    "\n",
    "validation_generator = test_datagen.flow_from_directory(\n",
    "    validation_data_dir,\n",
    "    target_size=(img_width, img_height),\n",
    "    batch_size=batch_size,\n",
    "    class_mode='categorical')\n",
    "\n",
    "history = model.fit_generator(\n",
    "    train_generator,\n",
    "    steps_per_epoch=nb_train_samples // batch_size,\n",
    "    epochs= 2,\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=nb_validation_samples // batch_size)\n",
    "\n",
    "model.save_weights('second_try.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}