the food network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow.contrib import layers\n",
    "import numpy as np\n",
    "import math\n",
    "import os\n",
    "from nbutil import imshow_multi\n",
    "%matplotlib inline\n",
    "from simple import flatten, avg\n",
    "import random\n",
    "from show_graph import show_graph\n",
    "from skimage import io\n",
    "from skimage.transform import resize\n",
    "import charttt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "IMAGE_SIZE = 64\n",
    "N_CLASSES = 102 # 101 foods + null class\n",
    "BATCH_SIZE = 32"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(?, 64, 64, 12)\n",
      "(?, 32, 32, 24)\n",
      "(?, 16, 16, 48)\n",
      "(?, 8, 8, 96)\n",
      "(?, 4, 4, 192)\n",
      "(?, 2, 2, 384)\n",
      "(?, 1, 1, 768)\n",
      "ok\n"
     ]
    }
   ],
   "source": [
    "# bias_init = tf.truncated_normal_initializer(mean=0.1, stddev=0.02)\n",
    "\n",
    "class Net(object):\n",
    "    def __init__(self, img_size, n_classes):\n",
    "        with tf.variable_scope('net'):\n",
    "            n_layers = int(math.log(img_size) / math.log(2))\n",
    "\n",
    "            inputs = tf.placeholder(tf.float32, [None, img_size, img_size, 3], name='inputs')\n",
    "            labels = tf.placeholder(tf.int64, [None], name='labels')\n",
    "            dropout = tf.placeholder(tf.float32, name='dropout')\n",
    "\n",
    "            initial_chans = 12\n",
    "            x = self.conv(inputs, initial_chans, 1)\n",
    "            print x.get_shape()\n",
    "\n",
    "            for i, layer in enumerate(xrange(n_layers)):\n",
    "                chans = initial_chans * 2 ** (i+1)\n",
    "                x = self.inception_module(x, chans, 'inception-{}'.format(i))\n",
    "                x = self.inception_module(x, chans, 'inception-{}-2'.format(i))\n",
    "                x = layers.batch_norm(x, is_training=True, updates_collections=None)\n",
    "                x = self.pool(x, 2, 2, type='avg')\n",
    "                if i == n_layers - 1:\n",
    "                    x = tf.nn.dropout(x, dropout)\n",
    "                print x.get_shape()\n",
    "                        \n",
    "            logits = self.conv(x, n_classes, 1, activation=tf.identity)\n",
    "            logits = tf.reshape(logits, [-1, n_classes])\n",
    "\n",
    "            loss = tf.reduce_sum(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=labels, logits=logits))\n",
    "            predictions = tf.argmax(logits, axis=-1)\n",
    "            accuracy = tf.reduce_mean(tf.cast(tf.equal(predictions, labels), tf.float32))\n",
    "\n",
    "            global_step = tf.contrib.framework.get_or_create_global_step()\n",
    "            lr = tf.placeholder(tf.float32, name='lr')\n",
    "            train_op = tf.train.AdamOptimizer(lr).minimize(loss, global_step=global_step)\n",
    "\n",
    "            self.inputs = inputs\n",
    "            self.labels = labels\n",
    "            self.loss = loss\n",
    "            self.logits = logits\n",
    "            self.predictions = predictions\n",
    "            self.accuracy = accuracy\n",
    "            self.global_step = global_step\n",
    "            self.train_op = train_op\n",
    "            self.lr = lr\n",
    "            self.dropout = dropout\n",
    "    \n",
    "    def conv(self, x, channels, ksize, activation=tf.nn.relu):\n",
    "        return layers.conv2d(x, \n",
    "                             channels,\n",
    "                             kernel_size=ksize, \n",
    "                             activation_fn=activation)\n",
    "    \n",
    "    def pool(self, x, ksize, stride, type='max'):\n",
    "        fn = {'max': tf.nn.max_pool, 'avg': tf.nn.avg_pool}[type]\n",
    "        return fn(x, [1, ksize, ksize, 1], [1, stride, stride, 1], 'SAME')\n",
    "    \n",
    "    def inception_module(self, x, out_channels, name):\n",
    "        # https://culurciello.github.io/tech/2016/06/04/nets.html\n",
    "        # out_channels channels must be a multiple of 4\n",
    "        \n",
    "        with tf.variable_scope(name):\n",
    "            channels = x.get_shape()[-1].value\n",
    "            bsize = out_channels / 4\n",
    "            \n",
    "            y1 = self.conv(x, bsize, 1)\n",
    "            y2 = self.conv(self.pool(x, 3, 1, 'avg'), bsize, 1)\n",
    "            y3 = self.conv(self.conv(x, bsize, 1), bsize, 3)\n",
    "            y4 = self.conv(self.conv(self.conv(x, bsize, 1), bsize, 3), bsize, 3)\n",
    "            \n",
    "            return tf.concat([y1, y2, y3, y4], axis=-1)\n",
    "\n",
    "n = Net(IMAGE_SIZE, N_CLASSES)\n",
    "print 'ok'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Inception module v3 \n",
    "\n",
    "<img src='https://culurciello.github.io/assets/nets/inceptionv3.jpg' width=300 />"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['apple_pie', 'baby_back_ribs', 'baklava', 'beef_carpaccio', 'beef_tartare', 'beet_salad', 'beignets', 'bibimbap', 'bread_pudding', 'breakfast_burrito', 'bruschetta', 'caesar_salad', 'cannoli', 'caprese_salad', 'carrot_cake', 'ceviche', 'cheese_plate', 'cheesecake', 'chicken_curry', 'chicken_quesadilla', 'chicken_wings', 'chocolate_cake', 'chocolate_mousse', 'churros', 'clam_chowder', 'club_sandwich', 'crab_cakes', 'creme_brulee', 'croque_madame', 'cup_cakes', 'deviled_eggs', 'donuts', 'dumplings', 'edamame', 'eggs_benedict', 'escargots', 'falafel', 'filet_mignon', 'fish_and_chips', 'foie_gras', 'french_fries', 'french_onion_soup', 'french_toast', 'fried_calamari', 'fried_rice', 'frozen_yogurt', 'garlic_bread', 'gnocchi', 'greek_salad', 'grilled_cheese_sandwich', 'grilled_salmon', 'guacamole', 'gyoza', 'hamburger', 'hot_and_sour_soup', 'hot_dog', 'huevos_rancheros', 'hummus', 'ice_cream', 'lasagna', 'lobster_bisque', 'lobster_roll_sandwich', 'macaroni_and_cheese', 'macarons', 'miso_soup', 'mussels', 'nachos', 'null', 'omelette', 'onion_rings', 'oysters', 'pad_thai', 'paella', 'pancakes', 'panna_cotta', 'peking_duck', 'pho', 'pizza', 'pork_chop', 'poutine', 'prime_rib', 'pulled_pork_sandwich', 'ramen', 'ravioli', 'red_velvet_cake', 'risotto', 'samosa', 'sashimi', 'scallops', 'seaweed_salad', 'shrimp_and_grits', 'spaghetti_bolognese', 'spaghetti_carbonara', 'spring_rolls', 'steak', 'strawberry_shortcake', 'sushi', 'tacos', 'takoyaki', 'tiramisu', 'tuna_tartare', 'waffles']\n"
     ]
    }
   ],
   "source": [
    "root = '../data/food-101/images'\n",
    "labels = sorted(os.listdir(root))\n",
    "\n",
    "paths_to_labels = {}\n",
    "\n",
    "for label in labels:\n",
    "    label_dir = os.path.join(root, label)\n",
    "    for name in os.listdir(label_dir):\n",
    "        if name.endswith('.jpg') or name.endswith('.png') or name.endswith('.gif'):\n",
    "            path = os.path.join(label_dir, name)\n",
    "            paths_to_labels[path] = label\n",
    "\n",
    "print labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def create_image_queue(filenames, batch_size=BATCH_SIZE, grayscale=False, size=128):\n",
    "    filename_tensor = tf.convert_to_tensor(filenames, dtype=tf.string)\n",
    "    filename_q = tf.train.slice_input_producer([filename_tensor], num_epochs=None, shuffle=True)[0]\n",
    "\n",
    "    image_255 = tf.image.decode_jpeg(tf.read_file(filename_q), channels=3)\n",
    "    image = tf.cast(image_255, tf.float32) / 255.0\n",
    "    if grayscale:\n",
    "        image = tf.image.grayscale_to_rgb(image)\n",
    "    \n",
    "    def resize_image(image):\n",
    "        pre_crop_size = tf.cast(size + tf.random_uniform([]) * size * 0.3, tf.int32)\n",
    "        image_as_batch_of_1 = tf.expand_dims(image, 0)\n",
    "        image = tf.image.resize_bilinear(image_as_batch_of_1, [pre_crop_size, pre_crop_size])[0]\n",
    "        image = tf.random_crop(image, [size, size, 3])\n",
    "        return image\n",
    "    \n",
    "    image = resize_image(image)\n",
    "\n",
    "    def distort_image(image):\n",
    "        # noise_amt = tf.abs(tf.random_normal([], stddev=0.2))\n",
    "        # distorted_image = image + tf.random_uniform([299, 299, 3], maxval=noise_amt)\n",
    "        distorted_image = tf.image.random_flip_left_right(image)\n",
    "        distorted_image = tf.image.random_brightness(distorted_image, max_delta=0.15)\n",
    "        distorted_image = tf.image.random_contrast(distorted_image, lower=0.8, upper=1.2)\n",
    "        return distorted_image\n",
    "    \n",
    "    image = distort_image(image)\n",
    "    \n",
    "    images_batch, filenames_batch = tf.train.shuffle_batch([image, filename_q], batch_size=batch_size, capacity=128, min_after_dequeue=64)\n",
    "    return images_batch, filenames_batch\n",
    "\n",
    "images_batch, filenames_batch = create_image_queue(paths_to_labels.keys(), size=IMAGE_SIZE)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# sess = tf.Session()\n",
    "# images_batch, filenames_batch = create_image_queue(paths_to_labels.keys(), size=IMAGE_SIZE)\n",
    "\n",
    "# init_op = tf.group(tf.global_variables_initializer(), tf.local_variables_initializer())\n",
    "# sess.run(init_op)\n",
    "# tf.train.start_queue_runners(sess=sess)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# show_graph(sess.graph)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# images, filenames = sess.run([images_batch, filenames_batch])\n",
    "# imshow_multi(list(images)[:5])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "9520: loss=118.897351583, acc=0.145833333333, epoch=3.01623762376\n",
      "9536: loss=104.361433983, acc=0.228515625, epoch=3.02130693069\n",
      "9552: loss=107.740716457, acc=0.234375, epoch=3.02637623762\n",
      "9568: loss=108.427424908, acc=0.201171875, epoch=3.03144554455\n",
      "9584: loss=110.983247757, acc=0.212890625, epoch=3.03651485149\n",
      "9600: loss=106.301193714, acc=0.208984375, epoch=3.04158415842\n",
      "9616: loss=108.37438345, acc=0.1953125, epoch=3.04665346535\n",
      "9632: loss=109.72287941, acc=0.19140625, epoch=3.05172277228\n",
      "9648: loss=109.233715534, acc=0.1953125, epoch=3.05679207921\n",
      "9664: loss=109.048866272, acc=0.18359375, epoch=3.06186138614\n",
      "9680: loss=106.629406929, acc=0.185546875, epoch=3.06693069307\n",
      "9696: loss=107.713367939, acc=0.224609375, epoch=3.072\n",
      "9712: loss=110.637801647, acc=0.181640625, epoch=3.07706930693\n",
      "9728: loss=106.35400486, acc=0.2109375, epoch=3.08213861386\n",
      "9744: loss=111.653268814, acc=0.19140625, epoch=3.08720792079\n",
      "9760: loss=108.195783138, acc=0.201171875, epoch=3.09227722772\n",
      "9776: loss=103.974459648, acc=0.224609375, epoch=3.09734653465\n",
      "9792: loss=106.811705589, acc=0.203125, epoch=3.10241584158\n",
      "9808: loss=107.325563431, acc=0.212890625, epoch=3.10748514851\n",
      "9824: loss=105.53068018, acc=0.21875, epoch=3.11255445545\n",
      "9840: loss=111.722161293, acc=0.193359375, epoch=3.11762376238\n",
      "9856: loss=108.324901581, acc=0.20703125, epoch=3.12269306931\n",
      "9872: loss=110.35079813, acc=0.19140625, epoch=3.12776237624\n",
      "9888: loss=106.372294426, acc=0.208984375, epoch=3.13283168317\n",
      "9904: loss=110.141825199, acc=0.212890625, epoch=3.1379009901\n",
      "9920: loss=110.859611988, acc=0.1953125, epoch=3.14297029703\n",
      "9936: loss=107.476271629, acc=0.1953125, epoch=3.14803960396\n",
      "9952: loss=111.230361462, acc=0.169921875, epoch=3.15310891089\n",
      "9968: loss=111.141862869, acc=0.17578125, epoch=3.15817821782\n",
      "9984: loss=105.46745348, acc=0.20703125, epoch=3.16324752475\n",
      "10000: loss=109.093006611, acc=0.201171875, epoch=3.16831683168\n",
      "10016: loss=109.182519913, acc=0.197265625, epoch=3.17338613861\n",
      "10032: loss=107.421440125, acc=0.205078125, epoch=3.17845544554\n",
      "10048: loss=111.89503336, acc=0.185546875, epoch=3.18352475248\n",
      "10064: loss=111.289087772, acc=0.17578125, epoch=3.18859405941\n",
      "10080: loss=107.412683487, acc=0.197265625, epoch=3.19366336634\n",
      "10096: loss=107.749298096, acc=0.181640625, epoch=3.19873267327\n",
      "10112: loss=109.189942837, acc=0.185546875, epoch=3.2038019802\n",
      "10128: loss=109.557487011, acc=0.189453125, epoch=3.20887128713\n",
      "10144: loss=110.939527035, acc=0.1796875, epoch=3.21394059406\n",
      "10160: loss=108.020443916, acc=0.20703125, epoch=3.21900990099\n",
      "10176: loss=108.956256866, acc=0.193359375, epoch=3.22407920792\n",
      "10192: loss=109.104433537, acc=0.16796875, epoch=3.22914851485\n",
      "10208: loss=111.355558872, acc=0.197265625, epoch=3.23421782178\n",
      "10224: loss=107.244224548, acc=0.181640625, epoch=3.23928712871\n",
      "10240: loss=104.664944172, acc=0.22265625, epoch=3.24435643564\n",
      "10256: loss=108.06652689, acc=0.212890625, epoch=3.24942574257\n",
      "10272: loss=109.713208675, acc=0.20703125, epoch=3.2544950495\n",
      "10288: loss=105.220573902, acc=0.21484375, epoch=3.25956435644\n",
      "10304: loss=107.437152863, acc=0.20703125, epoch=3.26463366337\n",
      "10320: loss=108.702053547, acc=0.181640625, epoch=3.2697029703\n",
      "10336: loss=108.954724312, acc=0.216796875, epoch=3.27477227723\n",
      "10352: loss=106.615551472, acc=0.208984375, epoch=3.27984158416\n",
      "10368: loss=108.167903423, acc=0.189453125, epoch=3.28491089109\n",
      "10384: loss=106.841363907, acc=0.201171875, epoch=3.28998019802\n",
      "10400: loss=108.989631176, acc=0.19921875, epoch=3.29504950495\n",
      "10416: loss=105.705545902, acc=0.24609375, epoch=3.30011881188\n",
      "10432: loss=106.456574917, acc=0.21875, epoch=3.30518811881\n",
      "10448: loss=111.253135681, acc=0.16796875, epoch=3.31025742574\n",
      "10464: loss=109.572598934, acc=0.181640625, epoch=3.31532673267\n",
      "10480: loss=106.501482487, acc=0.212890625, epoch=3.3203960396\n",
      "10496: loss=106.381530762, acc=0.201171875, epoch=3.32546534653\n",
      "10512: loss=105.589558125, acc=0.208984375, epoch=3.33053465347\n",
      "10528: loss=104.251642227, acc=0.2265625, epoch=3.3356039604\n",
      "10544: loss=106.888717175, acc=0.181640625, epoch=3.34067326733\n",
      "10560: loss=108.456616879, acc=0.23046875, epoch=3.34574257426\n",
      "10576: loss=107.494437695, acc=0.201171875, epoch=3.35081188119\n",
      "10592: loss=104.223457813, acc=0.24609375, epoch=3.35588118812\n",
      "10608: loss=107.740772724, acc=0.1953125, epoch=3.36095049505\n",
      "10624: loss=105.592549324, acc=0.203125, epoch=3.36601980198\n",
      "10640: loss=106.069452286, acc=0.19921875, epoch=3.37108910891\n",
      "10656: loss=110.386366367, acc=0.171875, epoch=3.37615841584\n",
      "10672: loss=107.682452679, acc=0.203125, epoch=3.38122772277\n",
      "10688: loss=108.051706791, acc=0.205078125, epoch=3.3862970297\n",
      "10704: loss=107.523375988, acc=0.18359375, epoch=3.39136633663\n",
      "10720: loss=107.166818142, acc=0.205078125, epoch=3.39643564356\n",
      "10736: loss=104.761672497, acc=0.2265625, epoch=3.4015049505\n",
      "10752: loss=105.710347652, acc=0.201171875, epoch=3.40657425743\n",
      "10768: loss=104.584311485, acc=0.21875, epoch=3.41164356436\n",
      "10784: loss=108.033557892, acc=0.2265625, epoch=3.41671287129\n",
      "10800: loss=111.165724277, acc=0.18359375, epoch=3.42178217822\n",
      "10816: loss=106.247496128, acc=0.1953125, epoch=3.42685148515\n",
      "10832: loss=109.971740723, acc=0.208984375, epoch=3.43192079208\n",
      "10848: loss=103.534250259, acc=0.21484375, epoch=3.43699009901\n",
      "10864: loss=104.232736588, acc=0.212890625, epoch=3.44205940594\n",
      "10880: loss=103.930848122, acc=0.20703125, epoch=3.44712871287\n",
      "10896: loss=106.006615162, acc=0.2109375, epoch=3.4521980198\n",
      "10912: loss=105.398620129, acc=0.20703125, epoch=3.45726732673\n",
      "10928: loss=109.080152512, acc=0.19921875, epoch=3.46233663366\n",
      "10944: loss=107.2240839, acc=0.20703125, epoch=3.46740594059\n",
      "10960: loss=108.598898411, acc=0.203125, epoch=3.47247524752\n",
      "10976: loss=108.016113758, acc=0.185546875, epoch=3.47754455446\n",
      "10992: loss=104.328825474, acc=0.23046875, epoch=3.48261386139\n",
      "11008: loss=106.673360825, acc=0.205078125, epoch=3.48768316832\n",
      "11024: loss=106.809914112, acc=0.203125, epoch=3.49275247525\n",
      "11040: loss=106.306480408, acc=0.21484375, epoch=3.49782178218\n",
      "11056: loss=106.550690174, acc=0.21484375, epoch=3.50289108911\n",
      "11072: loss=112.609418392, acc=0.171875, epoch=3.50796039604\n",
      "11088: loss=107.704820156, acc=0.224609375, epoch=3.51302970297\n",
      "11104: loss=107.467963219, acc=0.20703125, epoch=3.5180990099\n",
      "11120: loss=103.483485699, acc=0.224609375, epoch=3.52316831683\n",
      "11136: loss=104.100899696, acc=0.220703125, epoch=3.52823762376\n",
      "11152: loss=106.977602005, acc=0.19921875, epoch=3.53330693069\n",
      "11168: loss=109.124392509, acc=0.201171875, epoch=3.53837623762\n",
      "11184: loss=106.33075428, acc=0.1953125, epoch=3.54344554455\n",
      "11200: loss=106.696423531, acc=0.181640625, epoch=3.54851485149\n",
      "11216: loss=105.716939926, acc=0.236328125, epoch=3.55358415842\n",
      "11232: loss=107.501331806, acc=0.203125, epoch=3.55865346535\n",
      "11248: loss=109.373063087, acc=0.177734375, epoch=3.56372277228\n",
      "11264: loss=104.344650269, acc=0.19140625, epoch=3.56879207921\n",
      "11280: loss=102.008962154, acc=0.228515625, epoch=3.57386138614\n",
      "11296: loss=107.665771484, acc=0.19921875, epoch=3.57893069307\n",
      "11312: loss=106.646238327, acc=0.208984375, epoch=3.584\n",
      "11328: loss=106.879432678, acc=0.208984375, epoch=3.58906930693\n",
      "11344: loss=108.437748909, acc=0.201171875, epoch=3.59413861386\n",
      "11360: loss=109.231814861, acc=0.185546875, epoch=3.59920792079\n",
      "11376: loss=107.737100124, acc=0.21484375, epoch=3.60427722772\n",
      "11392: loss=111.018892288, acc=0.173828125, epoch=3.60934653465\n",
      "11408: loss=106.127933979, acc=0.205078125, epoch=3.61441584158\n",
      "11424: loss=111.271773815, acc=0.173828125, epoch=3.61948514851\n",
      "11440: loss=105.788783073, acc=0.212890625, epoch=3.62455445545\n",
      "11456: loss=107.098297596, acc=0.201171875, epoch=3.62962376238\n",
      "11472: loss=104.276355267, acc=0.2265625, epoch=3.63469306931\n",
      "11488: loss=109.257671833, acc=0.1953125, epoch=3.63976237624\n",
      "11504: loss=107.498432636, acc=0.208984375, epoch=3.64483168317\n",
      "11520: loss=109.597231388, acc=0.177734375, epoch=3.6499009901\n",
      "11536: loss=103.540791988, acc=0.248046875, epoch=3.65497029703\n",
      "11552: loss=105.837570667, acc=0.220703125, epoch=3.66003960396\n",
      "11568: loss=105.428026676, acc=0.216796875, epoch=3.66510891089\n",
      "11584: loss=107.903543472, acc=0.197265625, epoch=3.67017821782\n",
      "11600: loss=104.288894176, acc=0.21484375, epoch=3.67524752475\n",
      "11616: loss=108.067333698, acc=0.201171875, epoch=3.68031683168\n",
      "11632: loss=104.27975893, acc=0.2265625, epoch=3.68538613861\n",
      "11648: loss=107.942643166, acc=0.208984375, epoch=3.69045544554\n",
      "11664: loss=106.993809223, acc=0.212890625, epoch=3.69552475248\n",
      "11680: loss=103.182035446, acc=0.22265625, epoch=3.70059405941\n",
      "11696: loss=105.935938358, acc=0.216796875, epoch=3.70566336634\n",
      "11712: loss=108.987713337, acc=0.19140625, epoch=3.71073267327\n",
      "11728: loss=105.985341549, acc=0.234375, epoch=3.7158019802\n",
      "11744: loss=104.547306538, acc=0.212890625, epoch=3.72087128713\n",
      "11760: loss=107.370383739, acc=0.1875, epoch=3.72594059406\n",
      "11776: loss=107.694834709, acc=0.19140625, epoch=3.73100990099\n",
      "11792: loss=104.548215389, acc=0.228515625, epoch=3.73607920792\n",
      "11808: loss=108.232607841, acc=0.1875, epoch=3.74114851485\n",
      "11824: loss=108.112875462, acc=0.19140625, epoch=3.74621782178\n",
      "11840: loss=101.95853138, acc=0.244140625, epoch=3.75128712871\n",
      "11856: loss=104.774456501, acc=0.224609375, epoch=3.75635643564\n",
      "11872: loss=108.829209805, acc=0.201171875, epoch=3.76142574257\n",
      "11888: loss=106.733145714, acc=0.21875, epoch=3.7664950495\n",
      "11904: loss=105.481042385, acc=0.20703125, epoch=3.77156435644\n",
      "11920: loss=107.49555254, acc=0.21484375, epoch=3.77663366337\n",
      "11936: loss=103.691010952, acc=0.2265625, epoch=3.7817029703\n",
      "11952: loss=106.270393848, acc=0.1953125, epoch=3.78677227723\n",
      "11968: loss=104.702345848, acc=0.212890625, epoch=3.79184158416\n",
      "11984: loss=104.300115108, acc=0.19140625, epoch=3.79691089109\n",
      "12000: loss=106.595180988, acc=0.197265625, epoch=3.80198019802\n",
      "12016: loss=109.424120426, acc=0.21875, epoch=3.80704950495\n",
      "12032: loss=105.753093719, acc=0.212890625, epoch=3.81211881188\n",
      "12048: loss=106.001895428, acc=0.220703125, epoch=3.81718811881\n",
      "12064: loss=107.623087406, acc=0.1953125, epoch=3.82225742574\n",
      "12080: loss=102.850786686, acc=0.240234375, epoch=3.82732673267\n",
      "12096: loss=104.238276005, acc=0.240234375, epoch=3.8323960396\n",
      "12112: loss=104.967168808, acc=0.203125, epoch=3.83746534653\n",
      "12128: loss=102.953603745, acc=0.21875, epoch=3.84253465347\n",
      "12144: loss=105.210475922, acc=0.21875, epoch=3.8476039604\n",
      "12160: loss=105.42589283, acc=0.203125, epoch=3.85267326733\n",
      "12176: loss=108.281947136, acc=0.18359375, epoch=3.85774257426\n",
      "12192: loss=105.500183582, acc=0.201171875, epoch=3.86281188119\n",
      "12208: loss=105.943264484, acc=0.20703125, epoch=3.86788118812\n",
      "12224: loss=108.897413731, acc=0.208984375, epoch=3.87295049505\n",
      "12240: loss=103.201959133, acc=0.228515625, epoch=3.87801980198\n",
      "12256: loss=106.643307209, acc=0.205078125, epoch=3.88308910891\n",
      "12272: loss=108.080610275, acc=0.197265625, epoch=3.88815841584\n",
      "12288: loss=105.735094547, acc=0.20703125, epoch=3.89322772277\n",
      "12304: loss=108.649535179, acc=0.18359375, epoch=3.8982970297\n",
      "12320: loss=104.540406704, acc=0.22265625, epoch=3.90336633663\n",
      "12336: loss=105.112339973, acc=0.22265625, epoch=3.90843564356\n",
      "12352: loss=106.682454586, acc=0.2109375, epoch=3.9135049505\n",
      "12368: loss=103.62445116, acc=0.216796875, epoch=3.91857425743\n",
      "12384: loss=104.627499104, acc=0.197265625, epoch=3.92364356436\n",
      "12400: loss=103.62084198, acc=0.228515625, epoch=3.92871287129\n",
      "12416: loss=103.123727322, acc=0.234375, epoch=3.93378217822\n",
      "12432: loss=103.268662453, acc=0.228515625, epoch=3.93885148515\n",
      "12448: loss=108.009995937, acc=0.1953125, epoch=3.94392079208\n",
      "12464: loss=107.70930624, acc=0.205078125, epoch=3.94899009901\n",
      "12480: loss=103.342209339, acc=0.22265625, epoch=3.95405940594\n",
      "12496: loss=106.98790741, acc=0.208984375, epoch=3.95912871287\n",
      "12512: loss=109.939183235, acc=0.208984375, epoch=3.9641980198\n",
      "12528: loss=107.754480362, acc=0.1953125, epoch=3.96926732673\n",
      "12544: loss=105.318401814, acc=0.18359375, epoch=3.97433663366\n",
      "12560: loss=106.261040211, acc=0.20703125, epoch=3.97940594059\n",
      "12576: loss=106.723499298, acc=0.20703125, epoch=3.98447524752\n",
      "12592: loss=101.684524059, acc=0.23046875, epoch=3.98954455446\n",
      "12608: loss=104.923997879, acc=0.201171875, epoch=3.99461386139\n",
      "12624: loss=100.936253071, acc=0.1953125, epoch=3.99968316832\n",
      "12640: loss=103.022229195, acc=0.22265625, epoch=4.00475247525\n",
      "12656: loss=107.512854576, acc=0.19921875, epoch=4.00982178218\n",
      "12672: loss=100.902336597, acc=0.224609375, epoch=4.01489108911\n",
      "12688: loss=104.508121967, acc=0.2109375, epoch=4.01996039604\n",
      "12704: loss=101.958733082, acc=0.22265625, epoch=4.02502970297\n",
      "12720: loss=104.398694992, acc=0.208984375, epoch=4.0300990099\n",
      "12736: loss=108.005190372, acc=0.2109375, epoch=4.03516831683\n",
      "12752: loss=108.048283577, acc=0.220703125, epoch=4.04023762376\n",
      "12768: loss=103.252717018, acc=0.203125, epoch=4.04530693069\n",
      "12784: loss=105.519964695, acc=0.19921875, epoch=4.05037623762\n",
      "12800: loss=102.590615749, acc=0.228515625, epoch=4.05544554455\n",
      "12816: loss=109.075349331, acc=0.205078125, epoch=4.06051485149\n",
      "12832: loss=107.003806114, acc=0.2109375, epoch=4.06558415842\n",
      "12848: loss=104.005650043, acc=0.203125, epoch=4.07065346535\n",
      "12864: loss=105.403577805, acc=0.21484375, epoch=4.07572277228\n",
      "12880: loss=102.430084705, acc=0.244140625, epoch=4.08079207921\n",
      "12896: loss=105.688420773, acc=0.1953125, epoch=4.08586138614\n",
      "12912: loss=106.290719509, acc=0.185546875, epoch=4.09093069307\n",
      "12928: loss=102.176925182, acc=0.23828125, epoch=4.096\n",
      "12944: loss=105.108802319, acc=0.228515625, epoch=4.10106930693\n",
      "12960: loss=105.783894062, acc=0.224609375, epoch=4.10613861386\n",
      "12976: loss=105.995384216, acc=0.212890625, epoch=4.11120792079\n",
      "12992: loss=102.065803528, acc=0.2265625, epoch=4.11627722772\n",
      "13008: loss=108.409359932, acc=0.185546875, epoch=4.12134653465\n",
      "13024: loss=104.144742966, acc=0.20703125, epoch=4.12641584158\n",
      "13040: loss=106.357019424, acc=0.2109375, epoch=4.13148514851\n",
      "13056: loss=108.10406065, acc=0.201171875, epoch=4.13655445545\n",
      "13072: loss=107.933728218, acc=0.208984375, epoch=4.14162376238\n",
      "13088: loss=105.91738987, acc=0.193359375, epoch=4.14669306931\n",
      "13104: loss=105.7362113, acc=0.2109375, epoch=4.15176237624\n",
      "13120: loss=107.083346367, acc=0.197265625, epoch=4.15683168317\n",
      "13136: loss=102.788966179, acc=0.22265625, epoch=4.1619009901\n",
      "13152: loss=106.398667812, acc=0.220703125, epoch=4.16697029703\n",
      "13168: loss=109.744499207, acc=0.197265625, epoch=4.17203960396\n",
      "13184: loss=105.546580315, acc=0.2109375, epoch=4.17710891089\n",
      "13200: loss=107.949878216, acc=0.216796875, epoch=4.18217821782\n",
      "13216: loss=101.40644598, acc=0.220703125, epoch=4.18724752475\n",
      "13232: loss=102.80856657, acc=0.2109375, epoch=4.19231683168\n",
      "13248: loss=102.385956764, acc=0.240234375, epoch=4.19738613861\n",
      "13264: loss=105.305331707, acc=0.2109375, epoch=4.20245544554\n",
      "13280: loss=105.113852501, acc=0.212890625, epoch=4.20752475248\n",
      "13296: loss=107.907111645, acc=0.201171875, epoch=4.21259405941\n",
      "13312: loss=100.49384594, acc=0.234375, epoch=4.21766336634\n",
      "13328: loss=104.733706951, acc=0.232421875, epoch=4.22273267327\n",
      "13344: loss=104.427314281, acc=0.20703125, epoch=4.2278019802\n",
      "13360: loss=106.162277699, acc=0.22265625, epoch=4.23287128713\n",
      "13376: loss=106.025434017, acc=0.205078125, epoch=4.23794059406\n",
      "13392: loss=107.558464527, acc=0.208984375, epoch=4.24300990099\n",
      "13408: loss=106.281137943, acc=0.220703125, epoch=4.24807920792\n",
      "13424: loss=105.65037632, acc=0.185546875, epoch=4.25314851485\n",
      "13440: loss=105.735275269, acc=0.21875, epoch=4.25821782178\n",
      "13456: loss=102.498934746, acc=0.22265625, epoch=4.26328712871\n",
      "13472: loss=107.078559399, acc=0.203125, epoch=4.26835643564\n",
      "13488: loss=105.929699898, acc=0.224609375, epoch=4.27342574257\n",
      "13504: loss=105.91760397, acc=0.203125, epoch=4.2784950495\n",
      "13520: loss=105.116082668, acc=0.197265625, epoch=4.28356435644\n",
      "13536: loss=107.307970047, acc=0.203125, epoch=4.28863366337\n",
      "13552: loss=105.23841095, acc=0.205078125, epoch=4.2937029703\n",
      "13568: loss=106.279858589, acc=0.185546875, epoch=4.29877227723\n",
      "13584: loss=101.244163513, acc=0.2421875, epoch=4.30384158416\n",
      "13600: loss=101.80034399, acc=0.193359375, epoch=4.30891089109\n",
      "13616: loss=104.447577953, acc=0.205078125, epoch=4.31398019802\n",
      "13632: loss=106.197400093, acc=0.197265625, epoch=4.31904950495\n",
      "13648: loss=102.237011909, acc=0.23046875, epoch=4.32411881188\n",
      "13664: loss=103.957701683, acc=0.2265625, epoch=4.32918811881\n",
      "13680: loss=103.601019382, acc=0.2109375, epoch=4.33425742574\n",
      "13696: loss=104.806238174, acc=0.19921875, epoch=4.33932673267\n",
      "13712: loss=100.149731636, acc=0.2421875, epoch=4.3443960396\n",
      "13728: loss=107.123548985, acc=0.216796875, epoch=4.34946534653\n",
      "13744: loss=104.849176884, acc=0.220703125, epoch=4.35453465347\n",
      "13760: loss=103.061450005, acc=0.228515625, epoch=4.3596039604\n",
      "13776: loss=107.164720535, acc=0.19921875, epoch=4.36467326733\n",
      "13792: loss=101.285951614, acc=0.2578125, epoch=4.36974257426\n",
      "13808: loss=104.912708759, acc=0.244140625, epoch=4.37481188119\n",
      "13824: loss=105.831773281, acc=0.212890625, epoch=4.37988118812\n",
      "13840: loss=105.244476795, acc=0.23046875, epoch=4.38495049505\n",
      "13856: loss=107.66023159, acc=0.208984375, epoch=4.39001980198\n",
      "13872: loss=101.229393959, acc=0.240234375, epoch=4.39508910891\n",
      "13888: loss=101.317276001, acc=0.244140625, epoch=4.40015841584\n",
      "13904: loss=100.439595222, acc=0.224609375, epoch=4.40522772277\n",
      "13920: loss=101.593986034, acc=0.2421875, epoch=4.4102970297\n",
      "13936: loss=106.703489304, acc=0.236328125, epoch=4.41536633663\n",
      "13952: loss=106.26771307, acc=0.19921875, epoch=4.42043564356\n",
      "13968: loss=104.382554531, acc=0.22265625, epoch=4.4255049505\n",
      "13984: loss=105.585139275, acc=0.220703125, epoch=4.43057425743\n",
      "14000: loss=107.25272131, acc=0.212890625, epoch=4.43564356436\n",
      "14016: loss=106.033671856, acc=0.2265625, epoch=4.44071287129\n",
      "14032: loss=105.199351311, acc=0.232421875, epoch=4.44578217822\n",
      "14048: loss=101.912798405, acc=0.232421875, epoch=4.45085148515\n",
      "14064: loss=104.59972477, acc=0.205078125, epoch=4.45592079208\n",
      "14080: loss=102.300302505, acc=0.23046875, epoch=4.46099009901\n",
      "14096: loss=101.843874454, acc=0.2265625, epoch=4.46605940594\n",
      "14112: loss=102.437969685, acc=0.24609375, epoch=4.47112871287\n",
      "14128: loss=103.138967514, acc=0.22265625, epoch=4.4761980198\n",
      "14144: loss=108.484753132, acc=0.208984375, epoch=4.48126732673\n",
      "14160: loss=107.050991535, acc=0.216796875, epoch=4.48633663366\n",
      "14176: loss=104.158420086, acc=0.2265625, epoch=4.49140594059\n",
      "14192: loss=101.063166618, acc=0.21875, epoch=4.49647524752\n",
      "14208: loss=105.935740948, acc=0.197265625, epoch=4.50154455446\n",
      "14224: loss=105.367743492, acc=0.19921875, epoch=4.50661386139\n",
      "14240: loss=103.629919529, acc=0.208984375, epoch=4.51168316832\n",
      "14256: loss=103.825267792, acc=0.212890625, epoch=4.51675247525\n",
      "14272: loss=100.260672092, acc=0.23828125, epoch=4.52182178218\n",
      "14288: loss=104.929283142, acc=0.201171875, epoch=4.52689108911\n",
      "14304: loss=103.576216221, acc=0.251953125, epoch=4.53196039604\n",
      "14320: loss=106.826297283, acc=0.197265625, epoch=4.53702970297\n",
      "14336: loss=102.71372509, acc=0.236328125, epoch=4.5420990099\n",
      "14352: loss=104.025303364, acc=0.21484375, epoch=4.54716831683\n",
      "14368: loss=110.152932167, acc=0.189453125, epoch=4.55223762376\n",
      "14384: loss=109.203725338, acc=0.212890625, epoch=4.55730693069\n",
      "14400: loss=104.526883602, acc=0.21875, epoch=4.56237623762\n",
      "14416: loss=101.928326607, acc=0.20703125, epoch=4.56744554455\n",
      "14432: loss=102.610074997, acc=0.228515625, epoch=4.57251485149\n",
      "14448: loss=108.475413322, acc=0.208984375, epoch=4.57758415842\n",
      "14464: loss=104.626064301, acc=0.208984375, epoch=4.58265346535\n",
      "14480: loss=108.419334888, acc=0.193359375, epoch=4.58772277228\n",
      "14496: loss=103.20160532, acc=0.240234375, epoch=4.59279207921\n",
      "14512: loss=109.318490505, acc=0.19140625, epoch=4.59786138614\n",
      "14528: loss=105.318004131, acc=0.2109375, epoch=4.60293069307\n",
      "14544: loss=103.773090839, acc=0.228515625, epoch=4.608\n",
      "14560: loss=102.916946411, acc=0.228515625, epoch=4.61306930693\n",
      "14576: loss=106.49503231, acc=0.232421875, epoch=4.61813861386\n",
      "14592: loss=109.225497246, acc=0.201171875, epoch=4.62320792079\n",
      "14608: loss=108.47597599, acc=0.21875, epoch=4.62827722772\n",
      "14624: loss=102.960189819, acc=0.234375, epoch=4.63334653465\n",
      "14640: loss=105.514161587, acc=0.216796875, epoch=4.63841584158\n",
      "14656: loss=104.240258217, acc=0.2421875, epoch=4.64348514851\n",
      "14672: loss=101.640462399, acc=0.24609375, epoch=4.64855445545\n",
      "14688: loss=103.69275856, acc=0.22265625, epoch=4.65362376238\n",
      "14704: loss=102.019462109, acc=0.240234375, epoch=4.65869306931\n",
      "14720: loss=107.130321503, acc=0.201171875, epoch=4.66376237624\n",
      "14736: loss=98.9014987946, acc=0.2421875, epoch=4.66883168317\n",
      "14752: loss=106.372565746, acc=0.220703125, epoch=4.6739009901\n",
      "14768: loss=100.342723846, acc=0.24609375, epoch=4.67897029703\n",
      "14784: loss=102.615914822, acc=0.216796875, epoch=4.68403960396\n",
      "14800: loss=100.455807209, acc=0.26171875, epoch=4.68910891089\n",
      "14816: loss=107.537241459, acc=0.197265625, epoch=4.69417821782\n",
      "14832: loss=104.140791893, acc=0.22265625, epoch=4.69924752475\n",
      "14848: loss=105.661073208, acc=0.20703125, epoch=4.70431683168\n",
      "14864: loss=104.690165043, acc=0.21484375, epoch=4.70938613861\n",
      "14880: loss=103.600407124, acc=0.228515625, epoch=4.71445544554\n",
      "14896: loss=102.791220188, acc=0.236328125, epoch=4.71952475248\n",
      "14912: loss=101.674454212, acc=0.24609375, epoch=4.72459405941\n",
      "14928: loss=101.196764946, acc=0.20703125, epoch=4.72966336634\n",
      "14944: loss=104.812466145, acc=0.212890625, epoch=4.73473267327\n",
      "14960: loss=107.289090157, acc=0.201171875, epoch=4.7398019802\n",
      "14976: loss=104.256016731, acc=0.208984375, epoch=4.74487128713\n",
      "14992: loss=108.343610764, acc=0.1953125, epoch=4.74994059406\n",
      "15008: loss=103.545564651, acc=0.208984375, epoch=4.75500990099\n",
      "15024: loss=105.042626381, acc=0.228515625, epoch=4.76007920792\n",
      "15040: loss=103.116011143, acc=0.20703125, epoch=4.76514851485\n",
      "15056: loss=102.994826794, acc=0.22265625, epoch=4.77021782178\n",
      "15072: loss=102.21251297, acc=0.21875, epoch=4.77528712871\n",
      "15088: loss=102.117819309, acc=0.23828125, epoch=4.78035643564\n",
      "15104: loss=101.620384693, acc=0.21484375, epoch=4.78542574257\n",
      "15120: loss=100.24300766, acc=0.244140625, epoch=4.7904950495\n",
      "15136: loss=106.431708336, acc=0.19140625, epoch=4.79556435644\n",
      "15152: loss=102.05019474, acc=0.212890625, epoch=4.80063366337\n",
      "15168: loss=103.328612804, acc=0.216796875, epoch=4.8057029703\n",
      "15184: loss=102.281643391, acc=0.23828125, epoch=4.81077227723\n",
      "15200: loss=98.4697532654, acc=0.2578125, epoch=4.81584158416\n",
      "15216: loss=104.659281731, acc=0.216796875, epoch=4.82091089109\n",
      "15232: loss=105.16014719, acc=0.216796875, epoch=4.82598019802\n",
      "15248: loss=104.983385086, acc=0.228515625, epoch=4.83104950495\n",
      "15264: loss=106.692474365, acc=0.203125, epoch=4.83611881188\n",
      "15280: loss=106.4820261, acc=0.201171875, epoch=4.84118811881\n",
      "15296: loss=101.893937111, acc=0.23828125, epoch=4.84625742574\n",
      "15312: loss=100.320423126, acc=0.234375, epoch=4.85132673267\n",
      "15328: loss=106.040497303, acc=0.21484375, epoch=4.8563960396\n",
      "15344: loss=101.888199329, acc=0.220703125, epoch=4.86146534653\n",
      "15360: loss=103.514030457, acc=0.228515625, epoch=4.86653465347\n",
      "15376: loss=106.961339474, acc=0.20703125, epoch=4.8716039604\n",
      "15392: loss=103.504733562, acc=0.255859375, epoch=4.87667326733\n",
      "15408: loss=103.693712711, acc=0.228515625, epoch=4.88174257426\n",
      "15424: loss=103.914483547, acc=0.21484375, epoch=4.88681188119\n",
      "15440: loss=105.834354877, acc=0.203125, epoch=4.89188118812\n",
      "15456: loss=96.7570843697, acc=0.275390625, epoch=4.89695049505\n",
      "15472: loss=106.030345917, acc=0.220703125, epoch=4.90201980198\n",
      "15488: loss=102.15309906, acc=0.248046875, epoch=4.90708910891\n",
      "15504: loss=100.583577156, acc=0.2265625, epoch=4.91215841584\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-11-80a95d9c4f27>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     30\u001b[0m             \u001b[0mn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mimages\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     31\u001b[0m             \u001b[0mn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlabels\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mlabel_indices\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 32\u001b[0;31m             \u001b[0mn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdropout\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;36m0.5\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     33\u001b[0m         })\n\u001b[1;32m     34\u001b[0m         \u001b[0mlosses\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mloss\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    765\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    766\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 767\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    768\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    769\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    963\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    964\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m--> 965\u001b[0;31m                              feed_dict_string, options, run_metadata)\n\u001b[0m\u001b[1;32m    966\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    967\u001b[0m       \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_run\u001b[0;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1013\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1014\u001b[0m       return self._do_call(_run_fn, self._session, feed_dict, fetch_list,\n\u001b[0;32m-> 1015\u001b[0;31m                            target_list, options, run_metadata)\n\u001b[0m\u001b[1;32m   1016\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1017\u001b[0m       return self._do_call(_prun_fn, self._session, handle, feed_dict,\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m   1020\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1021\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1022\u001b[0;31m       \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1023\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1024\u001b[0m       \u001b[0mmessage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc\u001b[0m in \u001b[0;36m_run_fn\u001b[0;34m(session, feed_dict, fetch_list, target_list, options, run_metadata)\u001b[0m\n\u001b[1;32m   1002\u001b[0m         return tf_session.TF_Run(session, options,\n\u001b[1;32m   1003\u001b[0m                                  \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1004\u001b[0;31m                                  status, run_metadata)\n\u001b[0m\u001b[1;32m   1005\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1006\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_prun_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msession\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "chart = charttt.Board('foodnetwork').chart('foodnetwork 10 (64x64 input)')\n",
    "\n",
    "sv = tf.train.Supervisor(logdir='models/foodnetwork-10', save_model_secs=60 * 5)\n",
    "with sv.managed_session() as sess:\n",
    "    \n",
    "    # init_op = tf.group(tf.global_variables_initializer(), tf.local_variables_initializer())\n",
    "    # sess.run(init_op)\n",
    "    tf.train.start_queue_runners(sess=sess)\n",
    "    \n",
    "    losses = []\n",
    "    accs = []\n",
    "    \n",
    "    while not sv.should_stop():\n",
    "        step = sess.run(n.global_step)\n",
    "        images, filenames = sess.run([images_batch, filenames_batch])\n",
    "        label_strings = [paths_to_labels[path] for path in filenames]\n",
    "        label_indices = [labels.index(label) for label in label_strings]\n",
    "        \n",
    "        epoch = step * BATCH_SIZE * 1.0 / len(paths_to_labels)\n",
    "        lr = 0.001\n",
    "        if epoch > 0.1:\n",
    "            lr = 0.0005\n",
    "        if epoch > 1:\n",
    "            lr = 0.0001\n",
    "        if epoch > 3.2:\n",
    "            lr = 0.00005\n",
    "        \n",
    "        _, loss, acc = sess.run([n.train_op, n.loss, n.accuracy], feed_dict={\n",
    "            n.lr: lr,\n",
    "            n.inputs: images,\n",
    "            n.labels: label_indices,\n",
    "            n.dropout: 0.5\n",
    "        })\n",
    "        losses.append(loss)\n",
    "        accs.append(acc)\n",
    "        # print logits[0]\n",
    "        if step % 16 == 0:\n",
    "            chart.write(epoch, loss=avg(losses), accuracy=avg(accs))\n",
    "            print \"{}: loss={}, acc={}, epoch={}\".format(step, avg(losses), avg(accs), epoch)\n",
    "            losses = []\n",
    "            accs = []\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "# create a directory of 1000 random images:\n",
    "\n",
    "def create_null_dir():\n",
    "\n",
    "    import shutil\n",
    "\n",
    "    print nulldir\n",
    "\n",
    "    def iterate_images_recursively(path, extensions=['.jpeg', '.jpg', '.png', '.gif']):\n",
    "        def is_image(path):\n",
    "            for ext in extensions:\n",
    "                if path.lower().endswith(ext):\n",
    "                    return True\n",
    "            return False\n",
    "\n",
    "        for (dirpath, dirname, filenames) in os.walk(path):\n",
    "            for filename in filenames:\n",
    "                subpath = os.path.join(dirpath, filename)\n",
    "                if is_image(subpath):\n",
    "                    yield subpath\n",
    "\n",
    "    tiny_imagenet_images = list(iterate_images_recursively('../data/tiny-imagenet-200/train'))\n",
    "    floor_images = list(iterate_images_recursively('../data/floors'))\n",
    "\n",
    "    nulldir = os.path.join(root, 'null')\n",
    "    try:\n",
    "        os.rmdir(nulldir)\n",
    "    except OSError:\n",
    "        pass\n",
    "    os.mkdir(nulldir)\n",
    "\n",
    "    for i in xrange(1000):\n",
    "        path = random.choice(random.choice([tiny_imagenet_images, floor_images]))\n",
    "        ext = path.split('.')[-1]\n",
    "        name = os.path.join(nulldir, '{}.{}'.format(i, ext))\n",
    "        shutil.copyfile(path, name)\n",
    "\n",
    "# create_null_dir()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "! ls ../data/food-101/images/null"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def softmax(x):\n",
    "    return np.exp(x) / np.sum(np.exp(x), axis=0)\n",
    "\n",
    "with sv.managed_session() as sess:\n",
    "\n",
    "    img = io.imread('http://3.bp.blogspot.com/-agOrGBzjpKw/UHMB77A5NnI/AAAAAAAAD0Q/wOL7KXEh1-Y/s1600/DSC_0422.JPG')\n",
    "    img = resize(img, [IMAGE_SIZE, IMAGE_SIZE])\n",
    "    img = np.expand_dims(img, 0)\n",
    "    print img.shape\n",
    "\n",
    "    feed = {\n",
    "        n.dropout: 1,\n",
    "        n.inputs: img\n",
    "    }\n",
    "    predictions = softmax(sess.run(n.logits, feed_dict=feed)[0])\n",
    "    indices = sorted(range(102), key=lambda i: predictions[i], reverse=True)\n",
    "    for i in indices:\n",
    "        print labels[i], ':', predictions[i]\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "2.7.12+"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}