dog breed classification - ConvNet.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ConvNet - dont need feature vectors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "import numpy as np\n",
    "import cv2\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load Data - X, Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Load the training and testing data, X is images, y are labels\n",
    "BASE_IM_DIR = '/Users/Gabrielle/Kaggle/dog_breeds/train'\n",
    "\n",
    "img_path_list = [file_name for file_name in os.listdir(BASE_IM_DIR) if file_name.endswith('jpg')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "100\n",
      "200\n",
      "300\n",
      "400\n",
      "500\n",
      "600\n",
      "700\n",
      "800\n",
      "900\n",
      "1000\n",
      "1100\n",
      "1200\n",
      "1300\n",
      "1400\n",
      "1500\n",
      "1600\n",
      "1700\n",
      "1800\n",
      "1900\n",
      "2000\n",
      "2100\n",
      "2200\n",
      "2300\n",
      "2400\n",
      "2500\n",
      "2600\n",
      "2700\n",
      "2800\n",
      "2900\n",
      "3000\n",
      "3100\n",
      "3200\n",
      "3300\n",
      "3400\n",
      "3500\n",
      "3600\n",
      "3700\n",
      "3800\n",
      "3900\n",
      "4000\n",
      "4100\n",
      "4200\n",
      "4300\n",
      "4400\n",
      "4500\n",
      "4600\n",
      "4700\n",
      "4800\n",
      "4900\n",
      "5000\n",
      "5100\n",
      "5200\n",
      "5300\n",
      "5400\n",
      "5500\n",
      "5600\n",
      "5700\n",
      "5800\n",
      "5900\n",
      "6000\n",
      "6100\n",
      "6200\n",
      "6300\n",
      "6400\n",
      "6500\n",
      "6600\n",
      "6700\n",
      "6800\n",
      "6900\n",
      "7000\n",
      "7100\n",
      "7200\n",
      "7300\n",
      "7400\n",
      "7500\n",
      "7600\n",
      "7700\n",
      "7800\n",
      "7900\n",
      "8000\n",
      "8100\n",
      "8200\n",
      "8300\n",
      "8400\n",
      "8500\n",
      "8600\n",
      "8700\n",
      "8800\n",
      "8900\n",
      "9000\n",
      "9100\n",
      "9200\n",
      "9300\n",
      "9400\n",
      "9500\n",
      "9600\n",
      "9700\n",
      "9800\n",
      "9900\n",
      "10000\n",
      "10100\n",
      "10200\n"
     ]
    }
   ],
   "source": [
    "img_list = []\n",
    "for im_ind, im_file_path in enumerate(img_path_list):\n",
    "    if im_ind % 100 ==0:\n",
    "        print (im_ind)\n",
    "    # reading the image\n",
    "    img = cv2.imread(os.path.join(BASE_IM_DIR, im_file_path))\n",
    "    # converting from bgr to rgb\n",
    "    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
    "    # resizing\n",
    "    img = cv2.resize(img,(50,50))\n",
    "    # add dummy dimension\n",
    "    img_list.append(img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "labels_df = pd.read_csv('/Users/Gabrielle/Kaggle/dog_breeds/labels.csv') # id = image_name, breed = breed\n",
    "\n",
    "Y_list = []\n",
    "for im_path in img_path_list:\n",
    "    breed_label = str(labels_df[labels_df['id'] == im_path[:-4]]['breed'].values[0])\n",
    "    Y_list.append(breed_label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = np.array(img_list)\n",
    "Y = np.array(Y_list) # len(Y) == 10222"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Normalising images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = X.astype(np.float)\n",
    "X /= 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10222, 50, 50, 3) (10222,)\n"
     ]
    }
   ],
   "source": [
    "print(X.shape, Y.shape) # just checking. Dont need to add dummy dimensions because the photos are RGB (unlike the fasion MNIST greyscale)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "120\n"
     ]
    }
   ],
   "source": [
    "print(len(set(Y))) # No of unique labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## After running conv_model.fit, we got an error \n",
    "`ValueError: Error when checking target: expected activation_23 to have shape (None, 120) but got array with shape (7155, 1)`\n",
    "=> need to one hot encode our labels (120 in total)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10222, 120)\n"
     ]
    }
   ],
   "source": [
    "Y = pd.get_dummies(Y) # one-hot encoding\n",
    "print(Y.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Split X Y into train val test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    }
   ],
   "source": [
    "train_percent = 0.7\n",
    "val_percent = 0.2\n",
    "test_percent = 0.1\n",
    "\n",
    "total_samples = len(Y)\n",
    "\n",
    "train_index   = int(total_samples * train_percent)\n",
    "val_index     = int(total_samples * val_percent)\n",
    "test_index    = int(total_samples * test_percent)\n",
    "\n",
    "X_train = X[:train_index]\n",
    "X_val  = X[train_index: train_index + val_index]\n",
    "X_test  = X[train_index + val_index :]\n",
    "\n",
    "Y_train = Y[:train_index]\n",
    "Y_val   = Y[train_index: train_index + val_index]\n",
    "Y_test  = Y[train_index + val_index :]\n",
    "\n",
    "print(len(X_train) + len(X_val) + len(X_test) == len(X))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 7155 samples, validate on 2044 samples\n",
      "Epoch 1/100\n",
      "7155/7155 [==============================] - 24s 3ms/step - loss: 4.8134 - acc: 0.0147 - val_loss: 4.6808 - val_acc: 0.0342\n",
      "Epoch 2/100\n",
      "7155/7155 [==============================] - 19s 3ms/step - loss: 4.0678 - acc: 0.1248 - val_loss: 4.4969 - val_acc: 0.0509\n",
      "Epoch 3/100\n",
      "7155/7155 [==============================] - 19s 3ms/step - loss: 2.7813 - acc: 0.3895 - val_loss: 4.7404 - val_acc: 0.0558\n",
      "Epoch 4/100\n",
      "7155/7155 [==============================] - 19s 3ms/step - loss: 1.7104 - acc: 0.6274 - val_loss: 5.2511 - val_acc: 0.0587\n",
      "Epoch 5/100\n",
      "7155/7155 [==============================] - 20s 3ms/step - loss: 0.9946 - acc: 0.7976 - val_loss: 5.5713 - val_acc: 0.0572\n",
      "Epoch 6/100\n",
      "7155/7155 [==============================] - 20s 3ms/step - loss: 0.5822 - acc: 0.8889 - val_loss: 6.0059 - val_acc: 0.0509\n",
      "Epoch 00006: early stopping\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout, Activation\n",
    "from keras.layers.convolutional import Convolution2D\n",
    "from keras.layers.pooling import MaxPooling2D\n",
    "from keras.layers.core import  Flatten\n",
    "from keras.callbacks import EarlyStopping\n",
    "early_stop = EarlyStopping(monitor='val_loss', patience=4, verbose=1) \n",
    "\n",
    "conv_model1 = Sequential()\n",
    "conv_model1.add(Convolution2D(\n",
    "                        filters=20,\n",
    "                        kernel_size=3,\n",
    "                        activation='relu',\n",
    "                        padding='valid', # take care of padding for me\n",
    "                        input_shape=(50, 50,3)))\n",
    "conv_model1.add(Activation('relu'))\n",
    "conv_model1.add(MaxPooling2D( pool_size = (2,2)))\n",
    "conv_model1.add(Dropout(0.5))\n",
    "conv_model1.add(Flatten())\n",
    "conv_model1.add(Dense(120))\n",
    "conv_model1.add(Activation('softmax'))\n",
    "\n",
    "conv_model1.compile(loss='categorical_crossentropy', # NOT BINARY_CROSSENTROPY\n",
    "              optimizer='adam', # ADAM\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "history = conv_model1.fit(X_train, Y_train, batch_size=30, \n",
    "          epochs=100,verbose=1,\n",
    "          validation_data=(X_val, Y_val),\n",
    "          callbacks=[early_stop])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x14204cb70>"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAEWCAYAAABliCz2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\naJjSkQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1421590b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "### Well 0.8 accuracy sucks... probably because there are 120 labels. vs the other examples of CIFAR, MNIST - 10 labels only. Will have to train a better model"
   ]
  },
  {
   "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
}