model_expts.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload = 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using Theano backend.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using gpu device 0: GRID K520 (CNMeM is disabled)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential\n",
    "from keras.layers.convolutional import Convolution2D, MaxPooling2D\n",
    "from keras.layers.core import Dense, Dropout, Activation, Flatten\n",
    "from keras.optimizers import SGD\n",
    "import tables\n",
    "import os\n",
    "import theano"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## Loading lmdb dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import lmdb\n",
    "import caffe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "env = lmdb.open(\"/home/ubuntu/dataset/binary/DR_128_128_800_lmdb/train_db/\", readonly=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "with env.begin() as txn:\n",
    "    cursor = txn.cursor()\n",
    "    for key, val in cursor:\n",
    "        a = txn.get(key)\n",
    "        print type(a)\n",
    "        break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading h5 db"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_filename = \"data/train_128_db.h5\"\n",
    "test_filename = \"data/test_128_db.h5\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fp_train = tables.open_file(train_filename, mode='r+')\n",
    "X_train, y_train = fp_train.root.X_train, fp_train.root.y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fp_test = tables.open_file(test_filename, mode='r+')\n",
    "X_test, y_test = fp_test.root.X_test, fp_test.root.y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nn = Sequential()\n",
    "nn.add(Convolution2D(32, 3, 3, border_mode='same', input_shape=(3, 128, 128), activation='relu'))\n",
    "print \"Output shape after %d layers -\" % len(nn.layers), nn.output_shape\n",
    "nn.add(Convolution2D(32, 3, 3, activation='relu'))\n",
    "print \"Output shape after %d layers -\" % len(nn.layers), nn.output_shape\n",
    "nn.add(MaxPooling2D(pool_size=(3, 3)))\n",
    "print \"Output shape after %d layers -\" % len(nn.layers), nn.output_shape\n",
    "nn.add(Convolution2D(64, 3, 3, border_mode='same', activation='relu'))\n",
    "print \"Output shape after %d layers -\" % len(nn.layers), nn.output_shape\n",
    "nn.add(MaxPooling2D(pool_size=(3, 3)))\n",
    "print \"Output shape after %d layers -\" % len(nn.layers), nn.output_shape\n",
    "nn.add(Dropout(0.5))\n",
    "\n",
    "nn.add(Flatten())\n",
    "print \"Output shape after %d layers -\" % len(nn.layers), nn.output_shape\n",
    "\n",
    "nn.add(Dense(128, activation='relu'))\n",
    "nn.add(Dropout(0.5))\n",
    "print \"Output shape after %d layers -\" % len(nn.layers), nn.output_shape\n",
    "\n",
    "nn.add(Dense(5, activation='softmax'))\n",
    "\n",
    "sgd = SGD(lr=0.01, momentum=0.9, decay=1e-4, nesterov=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nn.compile(loss='binary_crossentropy', optimizer=sgd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch = np.random.randint(0, X_train.shape[0], size=16)\n",
    "print batch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nn.fit(X_train, y_train, batch_size=32, nb_epoch=20, show_accuracy=True, shuffle=True, validation_split=0.2, verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "fp_train.close()\n",
    "fp_test.close()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}