Image Completion.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Semantic Inpainting using DCGANs\n",
    "* * *\n",
    "\n",
    "This is an experimental tensorflow implementation of semantic inpainting from corrupted images using DCGANs from the paper [Semantic Image Inpainting with Perceptual and Contextual Losses](https://arxiv.org/abs/1607.07539). A major help was Brandon Amos blog on [Image Completion](https://bamos.github.io/2016/08/09/deep-completion/).One of the major difference between is the training method used.I have used Adam Optimizer instead of gradient descent.\n",
    "* * * *\n",
    "## Requirements\n",
    "* Tensorflow\n",
    "* glob\n",
    "* Python 3\n",
    "* * *\n",
    "## Dataset\n",
    "* I have used Celebrity faces dataset [CelebA](http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html).Download the aligned version ,extract in the same directory as the code.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import os,time\n",
    "from glob import glob\n",
    "\n",
    "from ops import batch_norm,linear,conv2d,deconv2d,lrelu\n",
    "from image_helpers import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model Defintion\n",
    "\n",
    "![alt text](images/model.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Parameter Defintion\n",
    "is_crop=True\n",
    "batch_size=64\n",
    "image_size=108\n",
    "sample_size=64\n",
    "image_shape=[64,64,3]\n",
    "\n",
    "z_dim=100\n",
    "\n",
    "gf_dim=64\n",
    "df_dim=64\n",
    "\n",
    "learning_rate=0.0002\n",
    "beta1=0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Batch Normalisation objects\n",
    "d_bn1 = batch_norm(name='d_bn1')\n",
    "d_bn2 = batch_norm(name='d_bn2')\n",
    "d_bn3 = batch_norm(name='d_bn3')\n",
    "\n",
    "g_bn0 = batch_norm(name='g_bn0')\n",
    "g_bn1 = batch_norm(name='g_bn1')\n",
    "g_bn2 = batch_norm(name='g_bn2')\n",
    "g_bn3 = batch_norm(name='g_bn3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def discriminator(image,reuse=False):\n",
    "    if reuse:\n",
    "        tf.get_variable_scope().reuse_variables()\n",
    "        \n",
    "    h0=lrelu(conv2d(image,df_dim,name='d_h0_conv'))\n",
    "    h1=lrelu(d_bn1(conv2d(h0,df_dim*2,name='d_h1_conv')))\n",
    "    h2=lrelu(d_bn2(conv2d(h1,df_dim*4,name='d_h2_conv')))\n",
    "    h3=lrelu(d_bn3(conv2d(h2,df_dim*8,name='d_h3_conv')))\n",
    "    h4=linear(tf.reshape(h3,[batch_size,-1]),1,'d_h3_lin')\n",
    "             \n",
    "    return tf.nn.sigmoid(h4),h4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def generator(z):\n",
    "    z_=linear(z,gf_dim*8*4*4,'g_h0_lin')\n",
    "    h0=tf.nn.relu(g_bn0(tf.reshape(z_,[-1,4,4,gf_dim*8])))\n",
    "    h1=tf.nn.relu(g_bn1(deconv2d(h0,[batch_size,8,8,gf_dim*4],name='g_h1')))\n",
    "    h2=tf.nn.relu(g_bn2(deconv2d(h1,[batch_size,16,16,gf_dim*2],name='g_h2')))\n",
    "    h3=tf.nn.relu(g_bn3(deconv2d(h2,[batch_size,32,32,gf_dim*1],name='g_h3')))\n",
    "    h4 = deconv2d(h3, [batch_size, 64, 64, 3], name='g_h4')\n",
    "    \n",
    "    return tf.nn.tanh(h4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Building model\n",
    "images=tf.placeholder(tf.float32,[batch_size]+image_shape,name='real_images')\n",
    "sample_images=tf.placeholder(tf.float32,[sample_size]+image_shape,name=\"sample_images\")\n",
    "z=tf.placeholder(tf.float32,[None,z_dim])\n",
    "\n",
    "G=generator(z)\n",
    "D,D_logits=discriminator(images)\n",
    "D_,D_logits_=discriminator(G,reuse=True)\n",
    "\n",
    "#cost fn\n",
    "d_loss_real=tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(D_logits,tf.ones_like(D)))\n",
    "d_loss_fake=tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(D_logits_,tf.zeros_like(D_)))\n",
    "d_loss=d_loss_real+d_loss_fake\n",
    "\n",
    "g_loss=tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(D_logits_,tf.ones_like(D_)))\n",
    "#For image completion\n",
    "mask=tf.placeholder(tf.float32,[None]+image_shape,name=\"mask\")\n",
    "\n",
    "contextual_loss=tf.reduce_sum(tf.contrib.layers.flatten(tf.abs(tf.mul(mask,G)-tf.mul(mask,images))))\n",
    "perceptual_loss=g_loss\n",
    "\n",
    "complete_loss=contextual_loss+perceptual_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#Optimizers\n",
    "t_vars=tf.trainable_variables()\n",
    "\n",
    "d_vars=[var for var in t_vars if 'd_' in var.name]\n",
    "g_vars=[var for var in t_vars if 'g_' in var.name]\n",
    "\n",
    "d_optim=tf.train.AdamOptimizer(learning_rate,beta1=beta1).minimize(d_loss,var_list=d_vars)\n",
    "g_optim=tf.train.AdamOptimizer(learning_rate,beta1=beta1).minimize(g_loss,var_list=g_vars)\n",
    "complete_optim=tf.train.AdamOptimizer(learning_rate,beta1=beta1).minimize(complete_loss,var_list=g_vars)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Points"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "sess=tf.Session()\n",
    "sess.run(tf.global_variables_initializer())\n",
    "\n",
    "saver=tf.train.Saver()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#DATASET \n",
    "data=glob(os.path.join('img_align_celeba/','*.jpg'))\n",
    "\n",
    "sample_z=np.random.uniform(-1,1,size=(sample_size,z_dim))\n",
    "sample_files=data[0:sample_size]\n",
    "sample=[get_image(sample_file,image_size,is_crop) for sample_file in sample_files]\n",
    "sample_images=np.reshape(np.array(sample).astype(np.float32),[sample_size]+image_shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 282487.0 7.2893 0.00100419\n",
      "2 276528.0 0.0993834 2.89487\n",
      "3 274097.0 1.12919 0.519172\n",
      "4 246674.0 1.34889 0.378413\n",
      "5 256940.0 0.111176 3.92779\n",
      "[Sample] d_loss: 0.11600902, g_loss: 5.41791153\n",
      "6 249439.0 1.29371 0.538651\n",
      "7 246668.0 0.28509 4.2912\n",
      "8 228919.0 1.40452 0.440909\n",
      "9 244363.0 0.73556 15.9906\n",
      "10 236065.0 0.0146455 12.1339\n",
      "[Sample] d_loss: 0.17829044, g_loss: 13.38587952\n",
      "11 248971.0 0.0783687 3.24196\n",
      "12 241467.0 4.29185 0.0225754\n",
      "13 239579.0 1.54853 26.4669\n",
      "14 242600.0 0.151014 25.2905\n",
      "15 220674.0 0.0228339 16.4789\n",
      "[Sample] d_loss: 0.09336887, g_loss: 18.01052094\n",
      "16 236997.0 0.0429531 5.2789\n",
      "17 227873.0 4.27723 0.0832719\n",
      "18 231394.0 1.18376 24.6325\n",
      "19 227557.0 0.474589 25.0789\n",
      "20 227269.0 0.0977529 16.9231\n",
      "[Sample] d_loss: 0.07977862, g_loss: 18.91727066\n",
      "21 229288.0 0.0904242 7.49572\n",
      "22 227796.0 2.97267 0.512974\n",
      "23 243101.0 0.312964 14.9833\n",
      "24 234674.0 0.685183 15.013\n",
      "25 244656.0 0.212154 7.82399\n",
      "[Sample] d_loss: 0.14891370, g_loss: 10.20151997\n",
      "26 228032.0 2.20547 0.859715\n",
      "27 218550.0 0.784921 13.7666\n",
      "28 224967.0 0.215524 8.14993\n",
      "29 233930.0 1.02506 1.54343\n",
      "30 221013.0 0.202823 5.29911\n",
      "[Sample] d_loss: 0.17596623, g_loss: 6.98355961\n",
      "31 233331.0 0.240918 4.31386\n",
      "32 221469.0 1.63559 0.867338\n",
      "33 243092.0 0.89673 11.5971\n",
      "34 232857.0 0.176128 4.57618\n",
      "35 233030.0 2.75673 0.224795\n",
      "[Sample] d_loss: 1.92172289, g_loss: 0.49400747\n",
      "36 216841.0 3.64848 21.3311\n",
      "37 226292.0 0.912832 18.7106\n",
      "38 237308.0 0.0992048 11.1966\n",
      "39 230912.0 0.0646105 3.83531\n",
      "40 221903.0 1.5059 0.402398\n",
      "[Sample] d_loss: 0.97079611, g_loss: 0.80373305\n",
      "41 220147.0 0.398372 13.7659\n",
      "42 233183.0 0.442206 14.4466\n",
      "43 221478.0 0.115688 8.85882\n",
      "44 227511.0 0.103556 3.29797\n",
      "45 225740.0 0.316684 1.58457\n",
      "[Sample] d_loss: 0.24566384, g_loss: 2.27758265\n",
      "46 221256.0 0.052991 6.1553\n",
      "47 218603.0 0.0858537 6.03024\n",
      "48 226705.0 0.102655 3.63272\n",
      "49 233635.0 0.358139 1.5445\n",
      "50 232327.0 0.150035 2.79923\n",
      "[Sample] d_loss: 0.53596234, g_loss: 3.24253750\n",
      "51 234514.0 0.233379 3.13415\n",
      "52 232832.0 0.26823 2.26006\n",
      "53 222596.0 0.330167 2.0539\n",
      "54 223363.0 0.299028 7.12743\n",
      "55 228575.0 0.0561099 7.84102\n",
      "[Sample] d_loss: 0.53555936, g_loss: 8.23306084\n",
      "56 227016.0 0.143323 4.58196\n",
      "57 230780.0 0.283975 2.77148\n",
      "58 230692.0 0.128337 8.43642\n",
      "59 217837.0 0.123963 3.59217\n",
      "60 234979.0 0.471197 1.31431\n",
      "[Sample] d_loss: 0.31983742, g_loss: 1.69221103\n",
      "61 232261.0 0.235006 13.6215\n",
      "62 229534.0 0.240922 12.777\n",
      "63 246775.0 0.0259529 10.1607\n",
      "64 232175.0 0.109241 6.28344\n",
      "65 223427.0 0.136213 3.65215\n",
      "[Sample] d_loss: 0.08389595, g_loss: 4.43322706\n",
      "66 231858.0 0.141451 3.11864\n",
      "67 233260.0 0.100814 7.95506\n",
      "68 229460.0 0.140937 8.82961\n",
      "69 228820.0 0.044231 6.68642\n",
      "70 231867.0 0.0391057 5.28831\n",
      "[Sample] d_loss: 0.06474630, g_loss: 5.89106798\n",
      "71 218485.0 0.0381183 4.97443\n",
      "72 240378.0 0.0414664 4.84758\n",
      "73 237909.0 0.0316792 4.8363\n",
      "74 225482.0 0.0500597 5.042\n",
      "75 229679.0 0.0375784 4.85953\n",
      "[Sample] d_loss: 0.06993507, g_loss: 5.25527048\n",
      "76 229876.0 0.0591319 3.37589\n",
      "77 243249.0 0.0781012 3.23792\n",
      "78 242992.0 0.0586319 3.61272\n",
      "79 226477.0 0.186252 2.37231\n",
      "80 227364.0 0.552157 11.5508\n",
      "[Sample] d_loss: 1.29886222, g_loss: 12.28591442\n",
      "81 234229.0 0.048188 9.57236\n",
      "82 240537.0 0.0134968 6.89563\n",
      "83 232499.0 0.0642662 3.8377\n",
      "84 233121.0 0.120957 2.86964\n",
      "85 242130.0 0.0207775 4.3802\n",
      "[Sample] d_loss: 0.05711756, g_loss: 5.11109400\n",
      "86 228492.0 0.0332307 4.22022\n",
      "87 233015.0 0.0767635 3.22753\n",
      "88 238375.0 0.09974 3.09681\n",
      "89 237020.0 0.304276 1.92144\n",
      "90 238323.0 0.262102 2.04917\n",
      "[Sample] d_loss: 0.36901647, g_loss: 2.30276966\n",
      "91 244171.0 0.314343 10.4446\n",
      "92 227545.0 0.0930861 10.2484\n",
      "93 234162.0 0.670985 0.967371\n",
      "94 239705.0 0.268768 7.71297\n",
      "95 249262.0 0.159713 3.97782\n",
      "[Sample] d_loss: 0.56368059, g_loss: 5.05822468\n",
      "96 247300.0 1.02076 2.36702\n",
      "97 230482.0 1.07567 13.0539\n",
      "98 250850.0 0.0403265 7.90645\n",
      "99 233118.0 0.209202 2.7842\n",
      "100 240843.0 0.362583 1.80479\n",
      "[Sample] d_loss: 0.40569168, g_loss: 1.82191539\n",
      "101 238528.0 0.073882 8.59026\n",
      "102 247490.0 0.154018 10.2599\n",
      "103 227374.0 0.0304791 8.15466\n",
      "104 237288.0 0.0117367 6.69026\n",
      "105 227734.0 0.0378877 5.03174\n",
      "[Sample] d_loss: 0.11802465, g_loss: 5.18348503\n",
      "106 233267.0 0.0735894 4.84565\n",
      "107 242341.0 0.0144858 6.22992\n",
      "108 238907.0 0.0219373 6.86812\n",
      "109 218899.0 0.042041 4.906\n",
      "110 231020.0 0.0303419 4.32999\n",
      "[Sample] d_loss: 0.05061100, g_loss: 5.01278400\n",
      "111 224129.0 0.0272198 4.95109\n",
      "112 215290.0 0.0161069 5.1942\n",
      "113 219428.0 0.0152109 5.16986\n",
      "114 228589.0 0.00783775 5.417\n",
      "115 226036.0 0.00965136 5.4604\n",
      "[Sample] d_loss: 0.05278186, g_loss: 5.65951967\n",
      "116 232838.0 0.00963528 5.43493\n",
      "117 229123.0 0.0242478 4.84256\n",
      "118 223836.0 0.0146208 4.93775\n",
      "119 229561.0 0.00900754 5.32163\n",
      "120 237225.0 0.019878 5.31551\n",
      "[Sample] d_loss: 0.09606962, g_loss: 5.70008469\n",
      "121 234889.0 0.00995109 5.78568\n",
      "122 230902.0 0.00981983 5.56307\n",
      "123 227099.0 0.0307825 4.4313\n",
      "124 223199.0 0.0304973 4.23355\n",
      "125 238231.0 0.0100031 5.27072\n",
      "[Sample] d_loss: 0.07589614, g_loss: 4.75178766\n",
      "126 232674.0 0.0109544 5.15324\n",
      "127 232122.0 0.0124768 5.62787\n",
      "128 239068.0 0.00577762 5.96274\n",
      "129 228096.0 0.0190791 5.0168\n",
      "130 227326.0 0.0173474 5.35883\n",
      "[Sample] d_loss: 0.08644515, g_loss: 5.01514482\n",
      "131 225977.0 0.0325693 4.46277\n",
      "132 216371.0 0.034312 5.03701\n",
      "133 222484.0 0.0163619 5.00208\n",
      "134 232366.0 0.0115113 5.22872\n",
      "135 241162.0 0.00764891 5.46183\n",
      "[Sample] d_loss: 0.07085460, g_loss: 5.79302406\n",
      "136 229574.0 0.0174105 5.45416\n",
      "137 226597.0 0.0170864 5.64671\n",
      "138 238275.0 0.0150278 5.16325\n",
      "139 233610.0 0.0112422 5.11028\n",
      "140 228893.0 0.0110473 5.22609\n",
      "[Sample] d_loss: 0.06534805, g_loss: 5.67438078\n",
      "141 236861.0 0.0140983 5.32736\n",
      "142 227393.0 0.0131771 5.90189\n",
      "143 235779.0 0.00848797 5.78246\n",
      "144 230493.0 0.00692687 5.88537\n",
      "145 242716.0 0.0110878 5.29196\n",
      "[Sample] d_loss: 0.08790626, g_loss: 5.80489397\n",
      "146 231225.0 0.0122205 5.11008\n",
      "147 238659.0 0.0117037 5.77645\n",
      "148 230638.0 0.0175756 5.61163\n",
      "149 220145.0 0.0278739 4.64126\n",
      "150 226944.0 0.0158838 4.76763\n",
      "[Sample] d_loss: 0.03432536, g_loss: 5.33688068\n",
      "151 232705.0 0.00881153 5.39108\n",
      "152 224537.0 0.0122822 5.68102\n",
      "153 228229.0 0.00735766 6.04768\n",
      "154 228388.0 0.00517893 5.75304\n",
      "155 238058.0 0.00882215 5.21119\n",
      "[Sample] d_loss: 0.04109722, g_loss: 5.27304268\n",
      "156 242456.0 0.013323 4.87104\n",
      "157 225957.0 0.01703 5.2169\n",
      "158 232924.0 0.00694908 5.58453\n",
      "159 231190.0 0.0120224 5.32643\n",
      "160 232224.0 0.0254301 5.13455\n",
      "[Sample] d_loss: 0.03752626, g_loss: 5.19616318\n",
      "161 233386.0 0.0114367 4.80176\n",
      "162 234069.0 0.0107298 4.98471\n",
      "163 235582.0 0.00837972 5.37866\n",
      "164 234374.0 0.0446169 5.53277\n",
      "165 245869.0 0.0129017 4.80986\n",
      "[Sample] d_loss: 0.04730193, g_loss: 5.91464090\n",
      "166 221256.0 0.0163708 5.28049\n",
      "167 229175.0 0.0199265 4.69881\n",
      "168 222870.0 0.0166284 4.85229\n",
      "169 232317.0 0.00842492 5.09982\n",
      "170 227150.0 0.0259353 4.66713\n",
      "[Sample] d_loss: 0.04824191, g_loss: 5.00397587\n",
      "171 233823.0 0.0238737 4.99047\n",
      "172 235648.0 0.0155936 5.40203\n",
      "173 238560.0 0.00894511 5.7128\n",
      "174 229837.0 0.0211374 4.93182\n",
      "175 239050.0 0.0333094 4.34906\n",
      "[Sample] d_loss: 0.11382335, g_loss: 4.91633129\n",
      "176 238777.0 0.023757 4.65059\n",
      "177 234549.0 0.0234066 4.3521\n",
      "178 243782.0 0.0352357 5.12954\n",
      "179 227239.0 0.17978 2.26344\n",
      "180 236638.0 0.0121121 5.23738\n",
      "[Sample] d_loss: 0.06577131, g_loss: 5.40220451\n",
      "181 251268.0 0.0172349 5.97721\n",
      "182 242239.0 0.036469 4.96705\n",
      "183 234031.0 0.0262188 5.1254\n",
      "184 221232.0 0.0773363 3.52267\n",
      "185 221862.0 0.0325552 6.59387\n",
      "[Sample] d_loss: 0.15868407, g_loss: 7.59856272\n",
      "186 243207.0 0.0648476 4.45383\n",
      "187 239173.0 0.027962 4.09254\n",
      "188 230761.0 0.0281782 4.12391\n",
      "189 239185.0 0.0126938 4.79153\n",
      "190 244615.0 0.0146423 5.00414\n",
      "[Sample] d_loss: 0.07880122, g_loss: 6.54589844\n",
      "191 248247.0 0.00847159 5.40055\n",
      "192 240515.0 0.026969 4.26755\n",
      "193 236745.0 0.0998925 2.85351\n",
      "194 237708.0 0.0573707 3.82132\n",
      "195 247277.0 0.0408176 3.58679\n",
      "[Sample] d_loss: 0.21264750, g_loss: 4.63467312\n",
      "196 246981.0 0.0785196 3.13357\n",
      "197 233499.0 3.76655 0.0827683\n",
      "198 237561.0 4.90307 30.8296\n",
      "199 250424.0 0.52979 26.5839\n",
      "200 234164.0 0.0186628 18.3486\n",
      "[Sample] d_loss: 0.15513425, g_loss: 17.83559799\n",
      "201 232475.0 0.0157238 6.64334\n",
      "202 234208.0 3.18346 0.227658\n",
      "203 227974.0 0.863236 19.1933\n",
      "204 240140.0 1.16967 21.0508\n",
      "205 233598.0 0.141638 13.3817\n",
      "[Sample] d_loss: 0.65094686, g_loss: 13.64598274\n",
      "206 232207.0 0.24277 8.10804\n",
      "207 234332.0 0.0454663 7.67026\n",
      "208 223944.0 0.0274578 6.72822\n",
      "209 223066.0 0.0218023 6.62184\n",
      "210 236240.0 0.0666431 5.78545\n",
      "[Sample] d_loss: 0.21489657, g_loss: 5.92019272\n",
      "211 226682.0 0.104861 5.09795\n",
      "212 228393.0 0.0354797 5.00646\n",
      "213 226480.0 0.049423 4.81701\n",
      "214 235252.0 0.0227832 4.63974\n",
      "215 224402.0 0.0704613 4.38986\n",
      "[Sample] d_loss: 0.14497535, g_loss: 4.62863350\n",
      "216 217486.0 0.0218949 4.87069\n",
      "217 216059.0 0.0339529 5.24329\n",
      "218 224107.0 0.0149704 5.64122\n",
      "219 218933.0 0.0127757 5.20033\n",
      "220 234791.0 0.00814328 5.82185\n",
      "[Sample] d_loss: 0.14541656, g_loss: 5.69943142\n",
      "221 228876.0 0.0135209 5.75629\n",
      "222 218893.0 0.0356815 4.68274\n",
      "223 234191.0 0.025902 4.10312\n",
      "224 228151.0 0.0417658 4.03452\n",
      "225 236395.0 0.0207397 4.40349\n",
      "[Sample] d_loss: 0.19967143, g_loss: 4.58780384\n",
      "226 233003.0 0.034534 4.67472\n",
      "227 235322.0 0.0396185 4.7476\n",
      "228 228168.0 0.0178029 4.64536\n",
      "229 227191.0 0.0482604 4.0262\n",
      "230 226153.0 0.0236118 4.64213\n",
      "[Sample] d_loss: 0.17000815, g_loss: 4.73817110\n",
      "231 230057.0 0.0196168 4.76789\n",
      "232 224689.0 0.0327076 4.1422\n",
      "233 236112.0 0.0594067 4.13156\n",
      "234 224057.0 0.0358456 4.3645\n",
      "235 243267.0 0.0194016 4.64152\n",
      "[Sample] d_loss: 0.16593489, g_loss: 4.59318209\n",
      "236 233658.0 0.0317861 4.4482\n",
      "237 233907.0 0.0294461 4.51846\n",
      "238 234301.0 0.0309488 4.22815\n",
      "239 229132.0 0.0273762 4.84214\n",
      "240 230384.0 0.034775 4.82667\n",
      "[Sample] d_loss: 0.21976697, g_loss: 4.76398659\n",
      "241 227761.0 0.0561302 4.00227\n",
      "242 227735.0 0.0388394 4.13304\n",
      "243 227644.0 0.0615722 4.99376\n",
      "244 219833.0 0.02863 5.09435\n",
      "245 229947.0 0.0252947 5.16666\n",
      "[Sample] d_loss: 0.16339926, g_loss: 5.42608356\n",
      "246 232544.0 0.0431259 4.42563\n",
      "247 231965.0 0.0561339 5.01132\n",
      "248 241051.0 0.022199 5.65404\n",
      "249 236994.0 0.0151434 5.73013\n",
      "250 224802.0 0.0376041 4.56005\n",
      "[Sample] d_loss: 0.08151779, g_loss: 4.96111298\n",
      "251 228488.0 0.0270349 4.38578\n",
      "252 233242.0 0.0187593 4.80222\n",
      "253 237141.0 0.0176837 5.26819\n",
      "254 236231.0 0.00765318 6.06183\n",
      "255 214626.0 0.039808 4.87397\n",
      "[Sample] d_loss: 0.04781735, g_loss: 5.29637814\n",
      "256 230600.0 0.0237795 4.79164\n",
      "257 242342.0 0.0140439 5.38596\n",
      "258 231235.0 0.0160225 5.6932\n",
      "259 224695.0 0.010161 6.27275\n",
      "260 226567.0 0.00903921 6.05928\n",
      "[Sample] d_loss: 0.05506280, g_loss: 6.80773401\n",
      "261 231709.0 0.0120723 6.1225\n",
      "262 236434.0 0.00530667 7.00091\n",
      "263 231345.0 0.0119392 6.38939\n",
      "264 232193.0 0.0170099 6.02789\n",
      "265 230570.0 0.0173669 6.03008\n",
      "[Sample] d_loss: 0.05970846, g_loss: 6.65830946\n",
      "266 241510.0 0.0116628 6.33305\n",
      "267 236238.0 0.0217201 5.94722\n",
      "268 225370.0 0.00849996 6.67379\n",
      "269 232619.0 0.00875042 6.60828\n",
      "270 232028.0 0.00872602 6.8441\n",
      "[Sample] d_loss: 0.04462141, g_loss: 6.80883694\n",
      "271 244737.0 0.0101465 6.41539\n",
      "272 233551.0 0.00834047 6.50659\n",
      "273 230720.0 0.0135347 6.46788\n",
      "274 228724.0 0.0076012 7.13802\n",
      "275 232588.0 0.0049731 6.90587\n",
      "[Sample] d_loss: 0.03411392, g_loss: 6.99442482\n",
      "276 238740.0 0.00439329 6.68919\n",
      "277 225808.0 0.00651315 6.39027\n",
      "278 230533.0 0.00664583 6.45124\n",
      "279 227631.0 0.00901968 6.08823\n",
      "280 244889.0 0.0068897 6.08488\n",
      "[Sample] d_loss: 0.05213916, g_loss: 6.06281948\n",
      "281 228423.0 0.0159711 5.71435\n",
      "282 232857.0 0.00987732 6.29269\n",
      "283 235713.0 0.00718818 5.89121\n",
      "284 229673.0 0.00825992 6.41971\n",
      "285 227430.0 0.00930632 6.52186\n",
      "[Sample] d_loss: 0.05738930, g_loss: 6.77016640\n",
      "286 222880.0 0.00964625 6.24607\n",
      "287 239041.0 0.00692302 6.66637\n",
      "288 226469.0 0.00916264 6.74069\n",
      "289 223526.0 0.0083958 7.05017\n",
      "290 222067.0 0.00745634 6.54768\n",
      "[Sample] d_loss: 0.04481876, g_loss: 6.89225626\n",
      "291 242108.0 0.00417067 6.56804\n",
      "292 227266.0 0.00855611 5.97746\n",
      "293 226375.0 0.00809982 5.86793\n",
      "294 224373.0 0.022894 5.58134\n",
      "295 223920.0 0.0123834 6.11103\n",
      "[Sample] d_loss: 0.04447897, g_loss: 6.30993271\n",
      "296 236501.0 0.00616622 5.95452\n",
      "297 239469.0 0.00486474 6.31477\n",
      "298 222471.0 0.01576 6.54476\n",
      "299 240584.0 0.00467833 6.55898\n",
      "300 238486.0 0.00528898 6.81481\n",
      "[Sample] d_loss: 0.03745259, g_loss: 7.19942045\n",
      "301 224714.0 0.00664556 6.7361\n",
      "302 226595.0 0.0116915 6.75714\n",
      "303 228254.0 0.00693907 6.5623\n",
      "304 234693.0 0.00505667 6.40266\n",
      "305 221397.0 0.00652016 5.97942\n",
      "[Sample] d_loss: 0.03662929, g_loss: 6.35684109\n",
      "306 229638.0 0.00842832 6.01297\n",
      "307 226824.0 0.0053036 6.23335\n",
      "308 233790.0 0.00348787 7.01063\n",
      "309 229991.0 0.00908346 6.22502\n",
      "310 230525.0 0.00385921 6.71605\n",
      "[Sample] d_loss: 0.05208379, g_loss: 7.07441711\n",
      "311 226536.0 0.00704353 6.57535\n",
      "312 227179.0 0.00547935 6.41332\n",
      "313 250212.0 0.00750827 5.82572\n",
      "314 234616.0 0.0117486 5.9731\n",
      "315 230237.0 0.0133015 6.64978\n",
      "[Sample] d_loss: 0.05998197, g_loss: 7.16182518\n",
      "316 228343.0 0.00534785 7.08888\n",
      "317 225581.0 0.0157808 6.26036\n",
      "318 236806.0 0.00949105 6.57611\n",
      "319 232100.0 0.00472928 7.41062\n",
      "320 231944.0 0.00337311 7.23025\n",
      "[Sample] d_loss: 0.02167877, g_loss: 7.32331753\n",
      "321 221511.0 0.00559661 7.0422\n",
      "322 223953.0 0.00968856 6.54792\n",
      "323 219441.0 0.0111878 6.31454\n",
      "324 225638.0 0.00953885 6.15228\n",
      "325 214925.0 0.00346153 6.94129\n",
      "[Sample] d_loss: 0.02705349, g_loss: 7.47104883\n",
      "326 223848.0 0.00326469 7.36999\n",
      "327 237949.0 0.00311648 6.71274\n",
      "328 236584.0 0.00306652 6.91049\n",
      "329 225408.0 0.00453587 6.55525\n",
      "330 231171.0 0.00934839 5.90724\n",
      "[Sample] d_loss: 0.03794486, g_loss: 6.75169849\n",
      "331 225152.0 0.012217 6.51665\n",
      "332 230556.0 0.00483899 6.39285\n",
      "333 230084.0 0.00272052 7.00029\n",
      "334 223528.0 0.00461105 6.46103\n",
      "335 224546.0 0.00504601 6.71855\n",
      "[Sample] d_loss: 0.03703976, g_loss: 7.29901505\n",
      "336 231152.0 0.00511401 6.85701\n",
      "337 239254.0 0.00500541 6.68827\n",
      "338 229720.0 0.00470691 6.61704\n",
      "339 230719.0 0.00571903 6.00766\n",
      "340 227765.0 0.00575892 7.02052\n",
      "[Sample] d_loss: 0.04576585, g_loss: 7.30345726\n",
      "341 222227.0 0.0097072 7.04605\n",
      "342 240589.0 0.00422425 6.60751\n",
      "343 231894.0 0.00696187 6.46882\n",
      "344 229089.0 0.00427696 6.55492\n",
      "345 229901.0 0.00593136 6.52651\n",
      "[Sample] d_loss: 0.04528400, g_loss: 7.29823494\n",
      "346 227550.0 0.00810985 6.94126\n",
      "347 225892.0 0.00799331 6.15584\n",
      "348 219861.0 0.00569321 6.8965\n",
      "349 227115.0 0.00653193 6.03038\n",
      "350 229437.0 0.00563783 6.22979\n",
      "[Sample] d_loss: 0.04784630, g_loss: 6.74322796\n",
      "351 230905.0 0.00641609 5.80035\n",
      "352 233912.0 0.00630432 5.98079\n",
      "353 242256.0 0.00503092 6.39165\n",
      "354 224562.0 0.00897856 6.77515\n",
      "355 240421.0 0.00533586 6.87426\n",
      "[Sample] d_loss: 0.06678400, g_loss: 7.56421757\n",
      "356 232685.0 0.00419753 6.74777\n",
      "357 223149.0 0.0120203 7.05257\n",
      "358 230422.0 0.00362152 7.04082\n",
      "359 228160.0 0.0116201 6.6046\n",
      "360 227745.0 0.00373293 6.82082\n",
      "[Sample] d_loss: 0.02574606, g_loss: 7.34196663\n",
      "361 230275.0 0.00525981 6.35709\n",
      "362 244093.0 0.00519018 6.48038\n",
      "363 235597.0 0.0039971 6.64699\n",
      "364 226262.0 0.00949493 6.13839\n",
      "365 236849.0 0.00457394 6.21822\n",
      "[Sample] d_loss: 0.03232419, g_loss: 6.53604078\n",
      "366 224912.0 0.00998274 5.41324\n",
      "367 221642.0 0.0158585 5.44584\n",
      "368 228092.0 0.0132993 5.72031\n",
      "369 231961.0 0.00926836 5.59224\n",
      "370 230446.0 0.0156366 5.76043\n",
      "[Sample] d_loss: 0.05859080, g_loss: 6.51921749\n",
      "371 229604.0 0.00957838 5.63284\n",
      "372 228373.0 0.018297 5.00355\n",
      "373 232860.0 0.0586759 3.77599\n",
      "374 249124.0 0.0452311 3.73321\n",
      "375 242459.0 0.0975956 2.91956\n",
      "[Sample] d_loss: 0.45042205, g_loss: 4.19774866\n",
      "376 229526.0 0.182092 2.59648\n",
      "377 234532.0 0.141931 7.53892\n",
      "378 247905.0 0.0364125 8.7273\n",
      "379 227781.0 0.0180163 6.67734\n",
      "380 244264.0 0.134298 4.32576\n",
      "[Sample] d_loss: 0.09217034, g_loss: 5.42544746\n",
      "381 236407.0 0.00794402 7.82291\n",
      "382 243881.0 0.00293721 9.8204\n",
      "383 232229.0 0.0115984 9.46745\n",
      "384 236722.0 0.0143918 7.44666\n",
      "385 242641.0 0.0121682 5.56875\n",
      "[Sample] d_loss: 0.03963409, g_loss: 7.38958311\n",
      "386 242344.0 0.0132331 5.31705\n",
      "387 238078.0 0.00671329 6.00698\n",
      "388 227904.0 0.013077 5.2371\n",
      "389 225122.0 0.00907737 5.66986\n",
      "390 225619.0 0.026194 5.22994\n",
      "[Sample] d_loss: 0.02320081, g_loss: 7.06579494\n",
      "391 235780.0 0.010019 5.65837\n",
      "392 230970.0 0.0260254 5.04433\n",
      "393 225584.0 0.0153327 5.01628\n",
      "394 231767.0 0.00903254 5.44734\n",
      "395 223767.0 0.0316662 5.07014\n",
      "[Sample] d_loss: 0.07280695, g_loss: 6.65995741\n",
      "396 233457.0 0.0105404 6.07634\n",
      "397 239050.0 0.0292042 4.57349\n",
      "398 227220.0 0.0166133 5.57333\n",
      "399 228879.0 0.0386154 4.38414\n",
      "400 230961.0 0.130367 5.91494\n",
      "[Sample] d_loss: 0.48303351, g_loss: 8.57436943\n",
      "401 251261.0 0.0485235 7.48811\n",
      "402 235816.0 0.0513761 5.30405\n",
      "403 231016.0 0.0499712 4.41322\n",
      "404 244146.0 0.0153289 5.27378\n",
      "405 238806.0 0.00900818 5.96332\n",
      "[Sample] d_loss: 0.15898013, g_loss: 7.82029533\n",
      "406 229355.0 0.0402518 4.73577\n",
      "407 241944.0 0.0303526 4.48466\n",
      "408 230988.0 0.0200604 4.89748\n",
      "409 248028.0 0.00590469 6.38302\n",
      "410 231242.0 0.0247886 5.31105\n",
      "[Sample] d_loss: 0.13028352, g_loss: 6.54356670\n",
      "411 245610.0 0.00773092 5.59995\n",
      "412 232449.0 0.0767833 3.69628\n",
      "413 239457.0 0.030232 4.92369\n",
      "414 245647.0 0.0838391 4.60188\n",
      "415 240851.0 0.0455272 4.64956\n",
      "[Sample] d_loss: 0.21393555, g_loss: 6.36150360\n",
      "416 241551.0 0.0351407 4.38862\n",
      "417 255684.0 0.0420684 5.30632\n",
      "418 238775.0 0.150728 2.48056\n",
      "419 235866.0 0.0259643 6.38019\n",
      "420 234560.0 0.0269685 7.99852\n",
      "[Sample] d_loss: 0.21031801, g_loss: 8.79819870\n",
      "421 231761.0 0.0224057 8.06307\n",
      "422 236712.0 0.0619987 5.24119\n",
      "423 239279.0 0.0307862 4.39839\n",
      "424 255368.0 0.525198 4.89987\n",
      "425 238345.0 2.88381 11.9246\n",
      "[Sample] d_loss: 2.85928965, g_loss: 13.85167885\n",
      "426 244172.0 0.248095 8.63135\n",
      "427 235127.0 0.291439 3.95281\n",
      "428 257956.0 0.0910518 9.64591\n",
      "429 252064.0 0.120828 12.2587\n",
      "430 227278.0 0.0268803 10.4078\n",
      "[Sample] d_loss: 0.01734102, g_loss: 10.02555561\n",
      "431 240899.0 0.00569455 7.07338\n",
      "432 225555.0 0.00434344 7.20843\n",
      "433 233632.0 0.00682019 5.91998\n",
      "434 231891.0 0.00571175 6.68227\n",
      "435 228809.0 0.00498165 7.40532\n",
      "[Sample] d_loss: 0.00722294, g_loss: 8.10242176\n",
      "436 235955.0 0.00644644 7.55073\n",
      "437 241113.0 0.00532984 6.94808\n",
      "438 230223.0 0.0135218 5.34129\n",
      "439 233142.0 0.0108817 5.3288\n",
      "440 228034.0 0.00952262 5.37383\n",
      "[Sample] d_loss: 0.01330117, g_loss: 5.76603603\n",
      "441 231402.0 0.00821999 5.51816\n",
      "442 227568.0 0.00408015 6.18706\n",
      "443 228192.0 0.00829474 5.68073\n",
      "444 218651.0 0.0237908 5.39958\n",
      "445 225750.0 0.0124995 4.77973\n",
      "[Sample] d_loss: 0.01857344, g_loss: 5.06382370\n",
      "446 235589.0 0.00758902 5.32547\n",
      "447 233694.0 0.0110533 6.06763\n",
      "448 226417.0 0.00624357 6.48237\n",
      "449 238758.0 0.0035881 6.20237\n",
      "450 240990.0 0.00392423 5.92283\n",
      "[Sample] d_loss: 0.01650183, g_loss: 6.27635145\n",
      "451 232324.0 0.00412346 5.84415\n",
      "452 242203.0 0.00412027 5.92891\n",
      "453 221197.0 0.0135743 6.02693\n",
      "454 226812.0 0.00651309 6.07469\n",
      "455 239609.0 0.00385881 6.0784\n",
      "[Sample] d_loss: 0.01295040, g_loss: 6.57402611\n",
      "456 225997.0 0.00427819 6.6498\n",
      "457 230986.0 0.00424453 6.12105\n",
      "458 227776.0 0.00432798 6.14206\n",
      "459 238730.0 0.00472092 6.11902\n",
      "460 241677.0 0.00357627 6.0541\n",
      "[Sample] d_loss: 0.01488798, g_loss: 6.25869131\n",
      "461 237075.0 0.0049489 5.72847\n",
      "462 244164.0 0.00539957 5.78004\n",
      "463 225896.0 0.00545449 6.13207\n",
      "464 234774.0 0.00411047 6.06359\n",
      "465 228435.0 0.00397428 6.38309\n",
      "[Sample] d_loss: 0.02083703, g_loss: 6.80359650\n",
      "466 229067.0 0.00358915 6.57241\n",
      "467 233783.0 0.00949546 6.33603\n",
      "468 233402.0 0.00234101 7.05441\n",
      "469 231492.0 0.00766266 6.60424\n",
      "470 239787.0 0.00391098 6.7012\n",
      "[Sample] d_loss: 0.01686109, g_loss: 6.90075827\n",
      "471 231652.0 0.00333893 6.43238\n",
      "472 219755.0 0.00497335 6.02615\n",
      "473 226189.0 0.0115621 5.88204\n",
      "474 234905.0 0.00633552 5.92082\n",
      "475 239756.0 0.00672242 5.72087\n",
      "[Sample] d_loss: 0.01657334, g_loss: 6.32542896\n",
      "476 238571.0 0.00530748 6.33439\n",
      "477 243502.0 0.00465349 6.63239\n",
      "478 232169.0 0.00376534 6.99403\n",
      "479 222980.0 0.00352129 7.08596\n",
      "480 236069.0 0.0031136 6.53782\n",
      "[Sample] d_loss: 0.01569846, g_loss: 7.41158009\n",
      "481 232192.0 0.00242 6.41409\n",
      "482 228513.0 0.0109052 6.61619\n",
      "483 234972.0 0.00871294 5.92578\n",
      "484 228919.0 0.00869686 5.4675\n",
      "485 233592.0 0.00493189 5.97714\n",
      "[Sample] d_loss: 0.00965434, g_loss: 6.47797871\n",
      "486 237099.0 0.00324196 6.09975\n",
      "487 234611.0 0.00283145 6.26349\n",
      "488 234069.0 0.00348271 6.39802\n",
      "489 231441.0 0.002699 6.30102\n",
      "490 241489.0 0.0023989 6.54204\n",
      "[Sample] d_loss: 0.01195450, g_loss: 7.13307381\n",
      "491 239964.0 0.0016464 6.68549\n",
      "492 230367.0 0.00231924 6.82873\n",
      "493 232170.0 0.00441406 6.88877\n",
      "494 230853.0 0.00745805 6.34587\n",
      "495 230504.0 0.00378106 6.78877\n",
      "[Sample] d_loss: 0.01029651, g_loss: 7.29813671\n",
      "496 235587.0 0.00184417 6.87682\n",
      "497 236726.0 0.0011928 7.37907\n",
      "498 240503.0 0.00184366 6.79634\n",
      "499 223160.0 0.00225777 6.55724\n",
      "500 241666.0 0.00305745 6.62385\n",
      "[Sample] d_loss: 0.01095224, g_loss: 6.89898205\n",
      "501 233296.0 0.00245379 6.40481\n",
      "502 235243.0 0.00330725 6.56732\n",
      "503 232420.0 0.00534135 6.49388\n",
      "504 241018.0 0.00165398 6.91947\n",
      "505 230548.0 0.00246502 7.16578\n",
      "[Sample] d_loss: 0.01184070, g_loss: 7.62761497\n",
      "506 234954.0 0.00104935 7.47584\n",
      "507 225812.0 0.0046939 7.36188\n",
      "508 233534.0 0.00153489 7.2661\n",
      "509 225678.0 0.0022783 7.1272\n",
      "510 230591.0 0.00539551 6.98715\n",
      "[Sample] d_loss: 0.01055986, g_loss: 7.12112188\n",
      "511 217440.0 0.00936802 6.90634\n",
      "512 232867.0 0.00261165 6.7012\n",
      "513 234356.0 0.00327527 6.8992\n",
      "514 236871.0 0.00230774 6.84718\n",
      "515 223101.0 0.00209308 7.12661\n",
      "[Sample] d_loss: 0.00679285, g_loss: 7.62021589\n",
      "516 229152.0 0.00168031 7.17106\n",
      "517 234952.0 0.0010182 7.75602\n",
      "518 232390.0 0.00121826 7.371\n",
      "519 221567.0 0.00198914 6.98283\n",
      "520 228544.0 0.00128378 7.25912\n",
      "[Sample] d_loss: 0.00779736, g_loss: 7.41330338\n",
      "521 232821.0 0.00174804 7.22761\n",
      "522 240278.0 0.00122231 7.24541\n",
      "523 223778.0 0.00375986 7.07968\n",
      "524 223159.0 0.00311893 6.90097\n",
      "525 235256.0 0.00259103 7.35689\n",
      "[Sample] d_loss: 0.00924281, g_loss: 7.59811592\n",
      "526 226965.0 0.00421429 7.35439\n",
      "527 230836.0 0.00373219 7.71545\n",
      "528 236153.0 0.00162868 7.6636\n",
      "529 224777.0 0.00116469 7.71395\n",
      "530 224451.0 0.00423772 7.50927\n",
      "[Sample] d_loss: 0.00709648, g_loss: 8.08582783\n",
      "531 225316.0 0.00303864 7.8389\n",
      "532 225797.0 0.00200309 7.36431\n",
      "533 238335.0 0.00148096 7.25667\n",
      "534 226151.0 0.00181822 7.06185\n",
      "535 238360.0 0.00425675 6.01023\n",
      "[Sample] d_loss: 0.00844006, g_loss: 6.44500446\n",
      "536 219348.0 0.00329617 6.7834\n",
      "537 224939.0 0.00290666 7.54935\n",
      "538 218319.0 0.00503667 7.64497\n",
      "539 218329.0 0.000921919 7.98389\n",
      "540 228556.0 0.00124708 8.06391\n",
      "[Sample] d_loss: 0.00554754, g_loss: 8.15587902\n",
      "541 224579.0 0.00116042 7.97102\n",
      "542 228131.0 0.0013109 7.55577\n",
      "543 229988.0 0.00165463 7.47293\n",
      "544 226960.0 0.00214207 7.31126\n",
      "545 234161.0 0.00198805 7.54495\n",
      "[Sample] d_loss: 0.00612833, g_loss: 7.87354469\n",
      "546 239810.0 0.00148918 7.38165\n",
      "547 233033.0 0.00209961 7.21658\n",
      "548 225831.0 0.00192223 7.40458\n",
      "549 229106.0 0.00158761 7.93108\n",
      "550 237197.0 0.00108034 7.80758\n",
      "[Sample] d_loss: 0.00687567, g_loss: 7.95410442\n",
      "551 235458.0 0.00150188 7.62764\n",
      "552 236869.0 0.0023154 7.18325\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-103-b6465b738f6f>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     29\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m     30\u001b[0m             \u001b[0mfd\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m{\u001b[0m\u001b[0mz\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mbatch_z\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mimages\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mbatch_images\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m---> 31\u001b[0;31m             \u001b[0msess\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mg_optim\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0md_optim\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mfd\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     32\u001b[0m             \u001b[0msess\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mcomplete_optim\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m{\u001b[0m\u001b[0mz\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mbatch_z\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mimages\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mbatch_images\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mmask\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mmask_\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m     33\u001b[0m             \u001b[0mc_loss\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mdloss\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mgloss\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0msess\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mcomplete_loss\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0md_loss\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mg_loss\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m{\u001b[0m\u001b[0mz\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mbatch_z\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mimages\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mbatch_images\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mmask\u001b[0m\u001b[1;33m:\u001b[0m\u001b[0mmask_\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mC:\\Users\\zc440z0ac\\AppData\\Local\\Continuum\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    764\u001b[0m     \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m    765\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 766\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    767\u001b[0m       \u001b[1;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m    768\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mC:\\Users\\zc440z0ac\\AppData\\Local\\Continuum\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    962\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[1;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m    963\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m--> 964\u001b[0;31m                              feed_dict_string, options, run_metadata)\n\u001b[0m\u001b[1;32m    965\u001b[0m     \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m    966\u001b[0m       \u001b[0mresults\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mC:\\Users\\zc440z0ac\\AppData\\Local\\Continuum\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\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   1012\u001b[0m     \u001b[1;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m   1013\u001b[0m       return self._do_call(_run_fn, self._session, feed_dict, fetch_list,\n\u001b[0;32m-> 1014\u001b[0;31m                            target_list, options, run_metadata)\n\u001b[0m\u001b[1;32m   1015\u001b[0m     \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m   1016\u001b[0m       return self._do_call(_prun_fn, self._session, handle, feed_dict,\n",
      "\u001b[0;32mC:\\Users\\zc440z0ac\\AppData\\Local\\Continuum\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m   1019\u001b[0m   \u001b[1;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m   1020\u001b[0m     \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1021\u001b[0;31m       \u001b[1;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1022\u001b[0m     \u001b[1;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m   1023\u001b[0m       \u001b[0mmessage\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0me\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mC:\\Users\\zc440z0ac\\AppData\\Local\\Continuum\\Anaconda3\\lib\\site-packages\\tensorflow\\python\\client\\session.py\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   1001\u001b[0m         return tf_session.TF_Run(session, options,\n\u001b[1;32m   1002\u001b[0m                                  \u001b[0mfeed_dict\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1003\u001b[0;31m                                  status, run_metadata)\n\u001b[0m\u001b[1;32m   1004\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m   1005\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_prun_fn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0msession\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "#Training\n",
    "counter=1\n",
    "start_time=time.time()\n",
    "for epoch in range(1):\n",
    "    np.random.shuffle(data)\n",
    "    batchidxs=int(len(data)/batch_size)\n",
    "    \n",
    "    for idx in range(batchidxs):\n",
    "        #try:\n",
    "            batch_files=data[idx*batch_size:(idx+1)*batch_size]\n",
    "            batch=[get_image(batch_file,image_size,is_crop=is_crop) for batch_file in batch_files]\n",
    "            batch_images=np.reshape(np.array(batch).astype(np.float32),[batch_size]+image_shape)\n",
    "\n",
    "            batch_z=np.random.uniform(-1,1,[batch_size,z_dim]).astype(np.float32)\n",
    "\n",
    "            #mask\n",
    "            scale=0.25\n",
    "            mask_=np.ones([batch_size]+image_shape).astype(np.float32)\n",
    "            l=int(64*scale)\n",
    "            u=int(64*(1.0-scale))\n",
    "            mask_[:,l:u,l:u,:]=0.0\n",
    "            #inverse mask\n",
    "            scale=0.25\n",
    "            imask_=np.zeros([batch_size]+image_shape).astype(np.float32)\n",
    "            l=int(64*scale)\n",
    "            u=int(64*(1.0-scale))\n",
    "            imask_[:,l:u,l:u,:]=1.0\n",
    "            \n",
    "            \n",
    "            fd={z:batch_z,images:batch_images}\n",
    "            sess.run([g_optim,d_optim],feed_dict=fd)\n",
    "            sess.run([complete_optim],feed_dict={z:batch_z,images:batch_images,mask:mask_})\n",
    "            c_loss,dloss,gloss=sess.run([complete_loss,d_loss,g_loss],feed_dict={z:batch_z,images:batch_images,mask:mask_})\n",
    "            print(counter,c_loss,dloss,gloss)\n",
    "            \n",
    "            if np.mod(counter,5)==0:\n",
    "                sample_generated,dl,gl=sess.run([G,d_loss,g_loss],feed_dict={z:sample_z,images:sample_images})\n",
    "                original_part=np.multiply(sample_images,mask_)\n",
    "                generated_part=np.multiply(sample_generated,imask_)\n",
    "                total=np.add(original_part,generated_part)\n",
    "                save_images(total,'samples\\\\')\n",
    "                print('[Sample] d_loss: %.8f, g_loss: %.8f' % (dl, gl))\n",
    "            counter+=1  \n",
    "        #except:\n",
    "         #   continue"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Saving the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'checkpoint\\\\image_completion.chk'"
      ]
     },
     "execution_count": 106,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saver.save(sess,\"checkpoint\\\\image_completion.chk\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "saver.restore(sess, \"checkpoint\\\\all_variables.chk\")\n",
    "print(sess.run(tf.all_variables()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [tensorflow]",
   "language": "python",
   "name": "Python [tensorflow]"
  },
  "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
}