Cats vs Dogs - attempt I.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Cats vs Dogs redux (kaggle)\n",
    "\n",
    "This code implements a deep learning classification model to distinguish between images of cats and images of dogs. It makes use of the pretrained VGG16 model<sup>1</sup> by using the Python `vgg` class [provided by Fast.AI](https://github.com/fastai/courses/blob/master/deeplearning1/nbs/vgg16.py) as part of the [Practical Deep Learning for Coders](http://course.fast.ai/lessons/lesson1.html) course as well as their [`utils.py`](https://github.com/fastai/courses/blob/master/deeplearning1/nbs/utils.py) package.\n",
    "\n",
    "The code assumes that their is a data folder whose structure is defined in the [Prepare Data](https://github.com/DanGolding/kaggle_cats_vs_dogs_redux/blob/master/Prepare%20data.ipynb) notebook in this repository.\n",
    "\n",
    "<sup>1</sup>K. Simonyan, A. Zisserman [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://arxiv.org/pdf/1409.1556.pdf), arXiv technical report, 2014"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using gpu device 0: Tesla K80 (CNMeM is disabled, cuDNN 5103)\n",
      "/home/ubuntu/anaconda2/lib/python2.7/site-packages/theano/sandbox/cuda/__init__.py:600: UserWarning: Your cuDNN version is more recent than the one Theano officially supports. If you see any problems, try updating Theano or downgrading cuDNN to version 5.\n",
      "  warnings.warn(warn)\n",
      "Using Theano backend.\n"
     ]
    }
   ],
   "source": [
    "from __future__ import division,print_function\n",
    "\n",
    "import os, json\n",
    "from glob import glob\n",
    "import numpy as np\n",
    "np.set_printoptions(precision=4, linewidth=100)\n",
    "from matplotlib import pyplot as plt\n",
    "from matplotlib.image import imread\n",
    "%matplotlib inline\n",
    "\n",
    "import utils; reload(utils)\n",
    "from utils import plots\n",
    "import vgg16; reload(vgg16)\n",
    "from vgg16 import Vgg16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "path = \"data/\"\n",
    "# path = \"data/sample/\"\n",
    "batch_size=64"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 23000 images belonging to 2 classes.\n",
      "Found 2000 images belonging to 2 classes.\n",
      "Epoch 1/1\n",
      "23000/23000 [==============================] - 1116s - loss: 0.1185 - acc: 0.9686 - val_loss: 0.0754 - val_acc: 0.9790\n"
     ]
    }
   ],
   "source": [
    "vgg = Vgg16()\n",
    "batches = vgg.get_batches(path+'train', batch_size=batch_size)\n",
    "val_batches = vgg.get_batches(path+'valid', batch_size=batch_size*2)\n",
    "vgg.finetune(batches)\n",
    "vgg.fit(batches, val_batches, nb_epoch=1)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "vgg.model.save_weights(path + 'results/ft1.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run the trained model on the test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 12500 images belonging to 1 classes.\n"
     ]
    }
   ],
   "source": [
    "test_batches, predictions = vgg.test(path + 'test', batch_size = batch_size*2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prepare the results for kaggle submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "filenames = test_batches.filenames\n",
    "utils.save_array(path + 'results/test_predictions.dat',predictions)\n",
    "utils.save_array(path + 'results/test_filenames.dat',filenames)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "ids = np.array([int(filename.split('.')[0].split(\"/\")[1]) for filename in filenames])\n",
    "results = np.vstack((ids,np.clip(predictions,0.05,0.95)[:,1])).T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Create the submission file and a link in order to download it to a local machine in order to submit to kaggle via the website"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "np.savetxt(path + 'results/results.csv',results,fmt='%d,%.5f',delimiter=',',header='id,label',comments='')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<a href='data/results/results.csv' target='_blank'>data/results/results.csv</a><br>"
      ],
      "text/plain": [
       "/home/ubuntu/FastAI/courses/deeplearning1/assignments/kaggle_cats_vs_dogs_redux/data/results/results.csv"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import FileLink\n",
    "FileLink(path + 'results/results.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "This submission scored 0.09907 on the kaggle competition. If the competition was still live, it would have placed 426/1315 i.e. in the top third."
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}