1 - Basic Convolutional Neural Network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "'''This script goes along the blog post\n",
    "\"Building powerful image classification models using very little data\"\n",
    "from blog.keras.io.\n",
    "It uses data that can be downloaded at:\n",
    "https://www.kaggle.com/c/dogs-vs-cats/data\n",
    "In our setup, we:\n",
    "- created a data/ folder\n",
    "- created train/ and validation/ subfolders inside data/\n",
    "- created cats/ and dogs/ subfolders inside train/ and validation/\n",
    "- put the cat pictures index 0-999 in data/train/cats\n",
    "- put the cat pictures index 1000-1400 in data/validation/cats\n",
    "- put the dogs pictures index 12500-13499 in data/train/dogs\n",
    "- put the dog pictures index 13500-13900 in data/validation/dogs\n",
    "So that we have 1000 training examples for each class, and 400 validation examples for each class.\n",
    "In summary, this is our directory structure:\n",
    "```\n",
    "data/\n",
    "    train/\n",
    "        dogs/\n",
    "            dog001.jpg\n",
    "            dog002.jpg\n",
    "            ...\n",
    "        cats/\n",
    "            cat001.jpg\n",
    "            cat002.jpg\n",
    "            ...\n",
    "    validation/\n",
    "        dogs/\n",
    "            dog001.jpg\n",
    "            dog002.jpg\n",
    "            ...\n",
    "        cats/\n",
    "            cat001.jpg\n",
    "            cat002.jpg\n",
    "            ...\n",
    "```\n",
    "'''\n",
    "\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 import backend as K\n",
    "\n",
    "# dimensions of our images.\n",
    "img_width, img_height = 150, 150\n",
    "\n",
    "train_data_dir = 'data/train'\n",
    "validation_data_dir = 'data/validation'\n",
    "nb_train_samples = 2000\n",
    "nb_validation_samples = 800\n",
    "epochs = 50\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)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bottleneck_fc_model.h5\n",
      "bottleneck_features_train.npy\n",
      "bottleneck_features_validation.npy\n",
      "classifier_from_little_data_script_1.out.txt\n",
      "classifier_from_little_data_script_1.py\n",
      "classifier_from_little_data_script_2.out.txt\n",
      "classifier_from_little_data_script_2.py\n",
      "classifier_from_little_data_script_3.out.txt\n",
      "classifier_from_little_data_script_3.py\n",
      "data\n",
      "first_try.h5\n",
      "launch.sub\n",
      "README.md\n",
      "requirements.txt\n",
      "summaries.py\n",
      "Untitled.ipynb\n",
      "vgg16_weights.h5\n",
      "words\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from subprocess import check_output\n",
    "print(check_output([\"ls\", \".\"]).decode(\"utf8\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_4 (Conv2D)            (None, 148, 148, 32)      896       \n",
      "_________________________________________________________________\n",
      "activation_6 (Activation)    (None, 148, 148, 32)      0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 74, 74, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 72, 72, 32)        9248      \n",
      "_________________________________________________________________\n",
      "activation_7 (Activation)    (None, 72, 72, 32)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 36, 36, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 34, 34, 64)        18496     \n",
      "_________________________________________________________________\n",
      "activation_8 (Activation)    (None, 34, 34, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 17, 17, 64)        0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 18496)             0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 64)                1183808   \n",
      "_________________________________________________________________\n",
      "activation_9 (Activation)    (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 1)                 65        \n",
      "_________________________________________________________________\n",
      "activation_10 (Activation)   (None, 1)                 0         \n",
      "=================================================================\n",
      "Total params: 1,212,513\n",
      "Trainable params: 1,212,513\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Conv2D(32, (3, 3), 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(64))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(1))\n",
    "model.add(Activation('sigmoid'))\n",
    "\n",
    "model.compile(loss='binary_crossentropy',\n",
    "              optimizer='rmsprop',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "print model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2000 images belonging to 2 classes.\n",
      "Found 800 images belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "# 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='binary')\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='binary')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "125/125 [==============================] - 143s 1s/step - loss: 0.6973 - acc: 0.5485 - val_loss: 0.6620 - val_acc: 0.6162\n",
      "Epoch 2/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.6487 - acc: 0.6330 - val_loss: 0.5987 - val_acc: 0.6837\n",
      "Epoch 3/50\n",
      "125/125 [==============================] - 14s 109ms/step - loss: 0.6122 - acc: 0.6830 - val_loss: 0.5921 - val_acc: 0.6837\n",
      "Epoch 4/50\n",
      "125/125 [==============================] - 13s 108ms/step - loss: 0.5804 - acc: 0.7030 - val_loss: 0.6243 - val_acc: 0.6750\n",
      "Epoch 5/50\n",
      "125/125 [==============================] - 14s 112ms/step - loss: 0.5679 - acc: 0.7115 - val_loss: 0.5546 - val_acc: 0.6875\n",
      "Epoch 6/50\n",
      "125/125 [==============================] - 14s 112ms/step - loss: 0.5473 - acc: 0.7270 - val_loss: 0.5235 - val_acc: 0.7462\n",
      "Epoch 7/50\n",
      "125/125 [==============================] - 13s 108ms/step - loss: 0.5469 - acc: 0.7395 - val_loss: 0.5655 - val_acc: 0.7163\n",
      "Epoch 8/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.5220 - acc: 0.7490 - val_loss: 0.5479 - val_acc: 0.7425\n",
      "Epoch 9/50\n",
      "125/125 [==============================] - 14s 111ms/step - loss: 0.5195 - acc: 0.7485 - val_loss: 0.5502 - val_acc: 0.7288\n",
      "Epoch 10/50\n",
      "125/125 [==============================] - 14s 113ms/step - loss: 0.5051 - acc: 0.7600 - val_loss: 0.5099 - val_acc: 0.7625\n",
      "Epoch 11/50\n",
      "125/125 [==============================] - 13s 108ms/step - loss: 0.4935 - acc: 0.7645 - val_loss: 0.5068 - val_acc: 0.7612\n",
      "Epoch 12/50\n",
      "125/125 [==============================] - 13s 108ms/step - loss: 0.4940 - acc: 0.7705 - val_loss: 0.5633 - val_acc: 0.7087\n",
      "Epoch 13/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.4743 - acc: 0.7885 - val_loss: 0.5504 - val_acc: 0.7575\n",
      "Epoch 14/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.4686 - acc: 0.7920 - val_loss: 0.5001 - val_acc: 0.7525\n",
      "Epoch 15/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.4563 - acc: 0.7875 - val_loss: 0.5038 - val_acc: 0.7475\n",
      "Epoch 16/50\n",
      "125/125 [==============================] - 14s 112ms/step - loss: 0.4485 - acc: 0.7930 - val_loss: 0.5095 - val_acc: 0.7762\n",
      "Epoch 17/50\n",
      "125/125 [==============================] - 14s 111ms/step - loss: 0.4422 - acc: 0.7945 - val_loss: 0.5289 - val_acc: 0.7700\n",
      "Epoch 18/50\n",
      "125/125 [==============================] - 14s 109ms/step - loss: 0.4316 - acc: 0.8010 - val_loss: 0.5616 - val_acc: 0.7525\n",
      "Epoch 19/50\n",
      "125/125 [==============================] - 14s 112ms/step - loss: 0.4221 - acc: 0.8085 - val_loss: 0.4811 - val_acc: 0.7900\n",
      "Epoch 20/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.4139 - acc: 0.8220 - val_loss: 0.4857 - val_acc: 0.7562\n",
      "Epoch 21/50\n",
      "125/125 [==============================] - 14s 109ms/step - loss: 0.4076 - acc: 0.8165 - val_loss: 0.5973 - val_acc: 0.7738\n",
      "Epoch 22/50\n",
      "125/125 [==============================] - 13s 108ms/step - loss: 0.4233 - acc: 0.8125 - val_loss: 0.5127 - val_acc: 0.7562\n",
      "Epoch 23/50\n",
      "125/125 [==============================] - 13s 105ms/step - loss: 0.4207 - acc: 0.8090 - val_loss: 0.8362 - val_acc: 0.7200\n",
      "Epoch 24/50\n",
      "125/125 [==============================] - 13s 108ms/step - loss: 0.4168 - acc: 0.8215 - val_loss: 0.6152 - val_acc: 0.7488\n",
      "Epoch 25/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.4151 - acc: 0.8145 - val_loss: 0.5806 - val_acc: 0.7425\n",
      "Epoch 26/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.4306 - acc: 0.8180 - val_loss: 0.4943 - val_acc: 0.7712\n",
      "Epoch 27/50\n",
      "125/125 [==============================] - 13s 105ms/step - loss: 0.4193 - acc: 0.8225 - val_loss: 0.5273 - val_acc: 0.7812\n",
      "Epoch 28/50\n",
      "125/125 [==============================] - 14s 111ms/step - loss: 0.4260 - acc: 0.8085 - val_loss: 0.4955 - val_acc: 0.7550\n",
      "Epoch 29/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.3994 - acc: 0.8300 - val_loss: 0.5761 - val_acc: 0.7863\n",
      "Epoch 30/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.4058 - acc: 0.8280 - val_loss: 0.6078 - val_acc: 0.7638\n",
      "Epoch 31/50\n",
      "125/125 [==============================] - 14s 109ms/step - loss: 0.4296 - acc: 0.8200 - val_loss: 0.4906 - val_acc: 0.7738\n",
      "Epoch 32/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.4246 - acc: 0.8200 - val_loss: 0.5235 - val_acc: 0.7625\n",
      "Epoch 33/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.4367 - acc: 0.8180 - val_loss: 0.5570 - val_acc: 0.8013\n",
      "Epoch 34/50\n",
      "125/125 [==============================] - 14s 108ms/step - loss: 0.4071 - acc: 0.8250 - val_loss: 0.5295 - val_acc: 0.7700\n",
      "Epoch 35/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.4109 - acc: 0.8310 - val_loss: 0.5368 - val_acc: 0.7913\n",
      "Epoch 36/50\n",
      "125/125 [==============================] - 14s 112ms/step - loss: 0.4115 - acc: 0.8255 - val_loss: 0.5052 - val_acc: 0.7775\n",
      "Epoch 37/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.4084 - acc: 0.8285 - val_loss: 0.6167 - val_acc: 0.7762\n",
      "Epoch 38/50\n",
      "125/125 [==============================] - 14s 109ms/step - loss: 0.4020 - acc: 0.8280 - val_loss: 0.5512 - val_acc: 0.7638\n",
      "Epoch 39/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.4104 - acc: 0.8280 - val_loss: 0.5622 - val_acc: 0.7475\n",
      "Epoch 40/50\n",
      "125/125 [==============================] - 14s 112ms/step - loss: 0.4107 - acc: 0.8345 - val_loss: 0.5083 - val_acc: 0.7400\n",
      "Epoch 41/50\n",
      "125/125 [==============================] - 13s 104ms/step - loss: 0.3931 - acc: 0.8320 - val_loss: 0.4750 - val_acc: 0.8100\n",
      "Epoch 42/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.3923 - acc: 0.8420 - val_loss: 0.6980 - val_acc: 0.7913\n",
      "Epoch 43/50\n",
      "125/125 [==============================] - 13s 107ms/step - loss: 0.3881 - acc: 0.8390 - val_loss: 0.6334 - val_acc: 0.7975\n",
      "Epoch 44/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.4019 - acc: 0.8255 - val_loss: 0.6023 - val_acc: 0.7650\n",
      "Epoch 45/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.3718 - acc: 0.8470 - val_loss: 0.5327 - val_acc: 0.7850\n",
      "Epoch 46/50\n",
      "125/125 [==============================] - 14s 109ms/step - loss: 0.3845 - acc: 0.8375 - val_loss: 0.5310 - val_acc: 0.7850\n",
      "Epoch 47/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.3827 - acc: 0.8435 - val_loss: 0.5491 - val_acc: 0.7913\n",
      "Epoch 48/50\n",
      "125/125 [==============================] - 14s 110ms/step - loss: 0.3763 - acc: 0.8420 - val_loss: 0.5367 - val_acc: 0.7937\n",
      "Epoch 49/50\n",
      "125/125 [==============================] - 13s 106ms/step - loss: 0.3931 - acc: 0.8370 - val_loss: 0.5441 - val_acc: 0.8000\n",
      "Epoch 50/50\n",
      "125/125 [==============================] - 14s 108ms/step - loss: 0.4094 - acc: 0.8325 - val_loss: 0.5628 - val_acc: 0.7825\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f409df04a10>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit_generator(\n",
    "    train_generator,\n",
    "    steps_per_epoch=nb_train_samples // batch_size,\n",
    "    epochs=epochs,\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=nb_validation_samples // batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Training a CNN from scratch, after 50 epochs, we have gotten an accuracy of ~78% in the validation set.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save_weights('first_try.h5')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Environment (conda_tensorflow_p27)",
   "language": "python",
   "name": "conda_tensorflow_p27"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}