VAE.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.examples.tutorials import mnist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From <ipython-input-2-5dba8e0e2d30>:1: read_data_sets (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n",
      "WARNING:tensorflow:From C:\\Users\\Dell\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:260: maybe_download (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please write your own downloading logic.\n",
      "WARNING:tensorflow:From C:\\Users\\Dell\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:262: extract_images (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tf.data to implement this functionality.\n",
      "Extracting ../../Neural networks/MNIST_data\\train-images-idx3-ubyte.gz\n",
      "WARNING:tensorflow:From C:\\Users\\Dell\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:267: extract_labels (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tf.data to implement this functionality.\n",
      "Extracting ../../Neural networks/MNIST_data\\train-labels-idx1-ubyte.gz\n",
      "Extracting ../../Neural networks/MNIST_data\\t10k-images-idx3-ubyte.gz\n",
      "Extracting ../../Neural networks/MNIST_data\\t10k-labels-idx1-ubyte.gz\n",
      "WARNING:tensorflow:From C:\\Users\\Dell\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:290: DataSet.__init__ (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n"
     ]
    }
   ],
   "source": [
    "data=mnist.input_data.read_data_sets(\"../../Neural networks/MNIST_data\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(55000, 784)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "images=data.train.images/255.\n",
    "images.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf.reset_default_graph()\n",
    "X=tf.placeholder(\"float32\",shape=(None,784),name=\"input\")\n",
    "Y=tf.placeholder(\"float32\",shape=(None,64),name=\"gen_input\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Model\n",
    "\n",
    "#encoder\n",
    "with tf.variable_scope(\"encoder\",reuse=tf.AUTO_REUSE):\n",
    "    L1=tf.keras.layers.Dense(units=784,activation=\"relu\",name=\"enc_input\")(X)\n",
    "    L2=tf.keras.layers.Dense(units=256,activation=\"relu\")(L1)\n",
    "            \n",
    "#encodings\n",
    "with tf.variable_scope(\"encoding\",reuse=tf.AUTO_REUSE):\n",
    "    encoding_mean=tf.keras.layers.Dense(64)(L2)\n",
    "    encoding_stddev=tf.keras.layers.Dense(64)(L2)\n",
    "    #sampling\n",
    "    std_norm=tf.random_normal(shape=(tf.shape(encoding_mean)[0], 64), mean=0, stddev=1)\n",
    "    Sample=encoding_mean + tf.exp(encoding_stddev) * std_norm\n",
    "     \n",
    "#Latent Vector\n",
    "with tf.variable_scope(\"Latent_vector\",reuse=tf.AUTO_REUSE):\n",
    "    Lt_vec=tf.keras.layers.Dense(64,name=\"Latent_vector\")(Sample)\n",
    "\n",
    "#decoding\n",
    "with tf.variable_scope(\"decoder\",reuse=tf.AUTO_REUSE):\n",
    "    dec_in=tf.keras.layers.Dense(256,activation=\"relu\",name=\"dec_input\")\n",
    "    dec_out=tf.keras.layers.Dense(784,activation=\"relu\",name=\"output\")\n",
    "    L3=dec_in(Lt_vec)\n",
    "    L4=dec_out(L3)\n",
    "\n",
    "#gen_output\n",
    "gen_dec=dec_in(Y)\n",
    "output=dec_out(gen_dec)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "#Loss\n",
    "out=L4\n",
    "\n",
    "rec_loss=tf.reduce_sum(tf.square(X-out))\n",
    "\n",
    "Kl_loss= 0.5 * tf.reduce_sum(tf.square(encoding_mean)+tf.square(tf.exp(encoding_stddev)) -1 -encoding_stddev)\n",
    "\n",
    "Loss= tf.reduce_mean(rec_loss + Kl_loss,name=\"net_loss\") \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "#optimizing\n",
    "\n",
    "opti=tf.train.AdamOptimizer(name=\"opti\")\n",
    "step=opti.minimize(Loss,name=\"step\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Session\n",
    "Sess=tf.Session()\n",
    "Sess.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1868894.1018066406\n",
      "1331646.4157714844\n",
      "1229359.8580322266\n",
      "1184940.6867675781\n",
      "1155878.3272705078\n"
     ]
    }
   ],
   "source": [
    "# Train\n",
    "batch_size = 100\n",
    "for i in range(5):\n",
    "    num_batches = int(data.train.num_examples/batch_size)\n",
    "    total_cost = 0\n",
    "    for j in range(num_batches):\n",
    "        batch_x, batch_y = data.train.next_batch(batch_size)\n",
    "        c, _ = Sess.run([Loss, step], feed_dict={X:batch_x})\n",
    "        total_cost += c\n",
    "    print(total_cost)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Generated_image')"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA2oAAADlCAYAAAAiJQSbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3XmcnFWV//HvSbqTkHQIhOyLJCQQFkXAAAIq+FMRN9wZeCEiwxAVMoIySkRHcBtxAQRBNAgERSPIIqsMDC4MOwRCEkgISQgQyMoaQghJ9/n98TwZivbe7q6uepZ0fd6vV17dfW7fuqcrdfqp089Tt8zdBQAAAAAoj15FJwAAAAAAeDMaNQAAAAAoGRo1AAAAACgZGjUAAAAAKBkaNQAAAAAoGRo1AAAAACgZGjUAAAAAUWb2BTO7owvf94qZ7ZBHTo2ARq2EzGypmb0/x/W+b2ZzzWyTmZ2e17pAkfKsMzMbZmYzzexZM3vJzO40s33zWBsoizxqzsxON7PLslwDqGRmh5vZvWa2zsxWpZ8fb2ZWdG6VzOzvZvZvWa/j7i3uviTrdRoFjRokaZGkb0i6sehEgB6qRdL9kt4habCkSyXdaGYthWYFAOg2MztZ0jmSfipphKThkr4k6QBJfXLMoymvtZAvGrUSS08z32lmZ5vZi2a2xMz2T+NPp3+5Obri+7czs+vN7GUzu9/MftCV09Tufqm7/0XS2kx/IKCE8qgzd1/i7me5+3J3b3X36UoO4pOy/vmAstl8CZWZ/czMXjCzJ8zsQxXjfzezH5nZfekZ6GvNbHA6dpCZLWt3e0vN7P1mdoikUyX9S3r51cP5/mRoJGY2SNL3JB3v7le6+1pPPOTuR7r7BjPrmz7OnzKzlWb2KzPbKp1/kJktM7OT0+PMcjM7puL2uzL3FDNbIekSM9vWzG4ws9VpXd1gZmPS7/+hpHdLOi+tjfPS+M5mdquZPW9mj5nZYRXrb2dm16XHuvskTeji/eJmNjH9fIaZ/dLM/pKue6eZjTCzn6c5LjCzPSvmTjOzxWa21sweNbNPVoz1NrMzzWxN+jtjarpW0+b/DzO7KL0fn0mPzb27979bHjRq5bevpDmStpP0B0l/lLS3pImSPqek6Db/Vf58SeuU/FXn6PQfgM7lWmdmtoeSRm1RzZkDW6Z9JT0maYikn0i6yOxNl4p9XtK/SholaZOkczu7QXe/WdJ/Sbo8vfzq7XXPGnjDfpL6Srq2g+/5saSdJO2h5HgyWtJ3KsZHSBqUxo+VdL6ZbVvF3MGStpc0Rclz+kvSr98iab2k8yTJ3b8l6X8lTU1rY6qZDZB0q5Jj3jBJR0j6pZntlt7++ZJekzRSSS3+axfvl/YOk/RtJbW+QdLdkh5Mv75S0lkV37tYSUM5SNJ3JV1mZiPTseMkfSi9P/aS9Il261yq5HfFREl7SjpYUuaXemaNRq38nnD3S9y9VdLlksZK+p67b3D3WyS9Lmli+leDT0s6zd1fdfdHlTxoAXQutzozs60l/U7Sd939pfr+GMAW40l3vzCtuUuVPBkcXjH+O3ef5+7rJP2npMN6wl/H0aMMkbTG3TdtDpjZXemVGevN7EAlzcVX3f15d1+r5A8Jh1fcxkYlx5qN7n6TpFckTUr/aNHZ3DYlx6IN7r7e3Z9z96vSY9NaST+UdGAH+X9U0tL02LfJ3R+UdJWkz1Qc677j7uvcfZ66/5zyGnef5e6vSbpG0mvu/tuK4+3/nVFz9z+5+7Pu3ubul0t6XNI+6fBhks5x92Xu/oKkMzbPM7PhSpq4k9J8V0k6u939tUXimtbyW1nx+XpJcvf2sRZJQ5X8fz5dMVb5OYC4XOosvWzlekn3uPuPup0tsOVbsfkTd381PZlW+ZrNyrp6UlKzkifGQFk8J2mImTVtbtbcfX9JSi/PHS6pv6RZFSeLTVLlHxyeq2z0JL2qN441nc1dnTY/Stfsr6Q5OUTS5rNyA82sd9oUtbe9pH3N7MWKWJOSPySGjnVPRu6HzrQ/loaOrZt/hs9L+pqkcWmoRW/U/SjFj73bK/kdsbzi/uqlHvA8mEat51it5JTvGEkL09jY4tIBeqRu15mZ9ZX0Z0nPSPpiJtkBPUdlXb1FyZmHNUouO+6/eSD9y//Qiu/1XLIDkkv4Nkj6uJIzUe2tUdKI7Obuz1R5212Z2/6xfrKS1z3v6+4r0kvsH1LS4IW+/2lJ/3D3D7S/4bSuNimpwwVp+C1V/gxVMbPtJV0o6X2S7nb3VjObrTfyX67k2LtZ5e+Ip5X8Xwxp1/hu8bj0sYdI/1pytaTTzay/me2s5Br/TplZs5n1U/J4aDKzflxiAvyz7taZmTUruRZ/vaTPu3tbtpkCW7zPmdmu6VmC70m6Mq2/hZL6mdlH0rr6tpLXCW22UtI4M+P5DTLl7i8qeR3VL83sM2bWYma90gZpgJJLEy+UdLaZDZMkMxttZh/swm13Z+5AJceYFy3ZfOe0duMrJVW+v9kNknYys6PS54HNZra3me0SONbtquz3PRigpJlcLUmWbKzy1orxKySdmN4P20g6ZfOAuy+XdIukM81s6/T/YUJ6+ekWjV9kPctUJS/AXKHk1PVMJX9h6MyFSor7CEnfSj8/KqMcgS1dd+psfyWvBzhYyUH0lfTfuzPNFNhy/U7SDCV11k/SVyQpfV3n8ZJ+o+Ts9DpJlbtA/in9+JyZPZhXsmhM7v4TJZfqfUPSKiXN0K+VNBF3pR8XSbrHzF6W9D/q+m6/1c79uaStlJyNu0fSze3Gz1Hy+rMXzOzc9HVsByt5HdezSmrtx3rjDx9TlVx6uEJJLV7Sxby7JX3N95lKzlSulPQ2SXdWfMuFSpqxOUrOFN6k5Kzf5ss6P69kk65HJb2g5I+jI7WFM3euEuipzOzHkka4O7s/AhmhzoD6MrO/S7rM3X9TdC4AysmSt/T4lbtvX3QuWeKMWg+Svh/G7pbYR8lWr9cUnRfQk1BnAADky8y2MrMPm1mTmY1Wcmlnjz/2splIzzJQyWVYo5Scgj9T0rXp5VV/CU1w95ZQHEAUdQYAaEgFHutMyWsCL1fyEp0b9eb3leuRuPQRAAAAAEqGSx8BAAAAoGRqatTM7BAze8zMFpnZtHolBeDNqDUge9QZkD3qDOi6bl/6mL7P1kJJH1CyNe79ko5It9cM6mN9vZ8GdGs9oKzW6oU17j608+/snmprjTpDT1S2OpOoNfRMWdYadQYkulpntWwmso+kRe6+RJLM7I9K3p09Wmz9NED72vtqWBIon//xK5/MeImqao06Q09UtjqTqDX0TBnXGnUGqOt1Vsulj6MlPV3x9bI0BqC+qDUge9QZkD3qDKhCLWfULBD7p+sozWyKpCmS1E/9a1gOaFid1hp1BtSMYxqQPeoMqEItZ9SWSRpb8fUYSc+2/yZ3n+7uk919crP61rAc0LA6rTXqDKgZxzQge9QZUIVaGrX7Je1oZuPNrI+kwyVdV5+0AFSg1oDsUWdA9qgzoArdvvTR3TeZ2VRJ/y2pt6SL3f2RumUGQBK1BuSBOgOyR50B1anlNWpy95sk3VSnXABEUGtA9qgzIHvUGdB1Nb3hNQAAAACg/mjUAAAAAKBkaNQAAAAAoGRo1AAAAACgZGjUAAAAAKBkatr1EQB6ulc/uW90bOIpjwbjK9cPjM5pfe8/vbcrAADAP+GMGgAAAACUDI0aAAAAAJQMjRoAAAAAlAyNGgAAAACUDI0aAAAAAJQMuz4CQAd+eOavo2Pjml4Jxo855sTonCax6yMAAOgcZ9QAAAAAoGRo1AAAAACgZGjUAAAAAKBkaNQAAAAAoGRo1AAAAACgZGjUAAAAAKBk2J4fQMPovd3g6Nj8H+8QjO/cfGd0zjuvPTkY3/G2e6tLDEBuVp2wf3Tsmyf+Phg/e/H7o3O2/tDimnMCgBDOqAEAAABAydCoAQAAAEDJ0KgBAAAAQMnQqAEAAABAydCoAQAAAEDJ1LTro5ktlbRWUqukTe4+uR5Jbel6Dx0aHVv6xR2D8fVjNlW9zhOHTo+ObfTWqm8v5tSV8f/WR/7f1sF464sv1W19UGv1svT4naNjiz50XjB+/4Z+0Tk7XLmx5pxQHtRZz/LEf+0XjN911E+jcwb1Ctf7d9b2j84JHwURQ53lw5r7ROLxp/6vfPBtwfgLk3pH57Q1h+N9X4jnts2i8LGz37Nr4+vMWRC/wR6sHtvzv9fd19ThdgB0jFoDskedAdmjzoAu4NJHAAAAACiZWhs1l3SLmc0ysyn1SAhAELUGZI86A7JHnQFdVOuljwe4+7NmNkzSrWa2wN1vr/yGtAinSFI/xa/xBtChDmuNOgPqgmMakD3qDOiims6oufuz6cdVkq6RtE/ge6a7+2R3n9ysvrUsBzSszmqNOgNqxzENyB51BnRdtxs1MxtgZgM3fy7pYEnz6pUYgAS1BmSPOgOyR50B1anl0sfhkq4xs8238wd3v7kuWRWgacTwYPz1HUdF56z42oZg/Ou73BKd8y8Db6ousQ5s9Hif3aa2uq3zg+H3RcfeddmRwfjgj7I9fx31qFrLQ9PIEcH4N468surbmjb1S9Gxvn+/v+rbQ2lRZ1sga4o/jTnyI/8IxmNb8EvS8cveE4xPmBbfNrz6N9dpaNRZTnrtOC4Yf/Ftg6NzVu9pwfhu+y2OzvnwsLnB+JqNA6Nz/rAo/I4M2/xhUHTOgDnRoR6t242auy+R9PY65gIggFoDskedAdmjzoDqsD0/AAAAAJQMjRoAAAAAlAyNGgAAAACUDI0aAAAAAJRMrW94vUVpGh3fwfHRb48Jxhccen50Tq9In/vT53aNzvlsZKebpy/fITqn/5rwDo6tfaJTdNi08M6TJ2z7WHxSN/i129X19oB6WP7r8M5RRw1cEZ1z8/rwm6oOePCp6Bx2ewOK9fhPw8dUSbp+SPz4HXP7LbsH4+OW3F31bQFFan10YTC+9ePxJ49Nr+0RjM9pGRedM2bfF4PxNa8PiM7Z+MjWwfiAK6mz9jijBgAAAAAlQ6MGAAAAACVDowYAAAAAJUOjBgAAAAAlQ6MGAAAAACVDowYAAAAAJdNQ2/MvOz+8ZbckLdg7vI3vuS/sHJ1zy5feHYz3WbQ8OmfTivD24MMU3zY8ZtlVu0XH6rkN/+6XfCU6Nu5CtlJFMZrGht9SQ5KmTLyj6tv76VeOCsb7rri/6tsCkI++Y1+pes5Gb42ODZnrtaQDlJ5vfD06NuDWecF4y8Twtv2S9N+DdwnGN61rjs7Z5bzFwXi8MhsXZ9QAAAAAoGRo1AAAAACgZGjUAAAAAKBkaNQAAAAAoGRo1AAAAACgZBpq18fx2z4fHYvtbDjhzAXROb1emB2Mb6ouLUmS7/f26NjRM24Ixg9rmRWd0xaJX/PKsOici79waDA+7t77onOArDWNHhWMj/jTi9E5xw16Ohj/2MKPRuf0u21OMM4ecEB59euzseo5p63aNzrWcsU9taQDbNFa99gxGF87If7MdvTgl4PxTTfFn2+2rlxVXWINjDNqAAAAAFAyNGoAAAAAUDI0agAAAABQMjRqAAAAAFAyNGoAAAAAUDKdNmpmdrGZrTKzeRWxwWZ2q5k9nn7cNts0gZ6PWgOyR50B2aPOgProyvb8MySdJ+m3FbFpkm5z9zPMbFr69Sn1T6++1h+4Mjo2TuGx1m6s03vo0OjY/O+PD8YXfuyCqtdpi27CL/38+V2D8X8cvld0jj3ycNU5oK5mqIfUWj1t2HFEMD59bPhtKzqy+K7to2PjNjxb9e1hizRD1NkWp/duk4Lx+98xMzondoS8+pE9onMm6qFq0kLcDFFnpbT+E/tEx4advCQYP21o+O1rJOm7d4Tf2mmny3iri3ro9Iyau98uqf0bkH1c0qXp55dK+kSd8wIaDrUGZI86A7JHnQH10d3XqA139+WSlH6Mv6sdgFpQa0D2qDMge9QZUKWuXPpYEzObImmKJPVT/6yXAxoSdQbkg1oDskedAYnunlFbaWYjJSn9uCr2je4+3d0nu/vkZvXt5nJAw+pSrVFnQE04pgHZo86AKnW3UbtO0tHp50dLurY+6QBoh1oDskedAdmjzoAqdXrpo5nNlHSQpCFmtkzSaZLOkHSFmR0r6SlJn80yybKyvuG/8mz8Y7/onAU7nx+Mx/dvjPv04+GddiRp08mDg3F/5JFurIQ8UGv18/u14Zc+TLxoeXTOpqySQalQZ1umx75Zv8vfWh7Yqm63hTDqLB9N4+M7GS85enQw/t6PPBid84tRdwXjN77aEp2z3d3N0THUrtNGzd2PiAy9r865AA2NWgOyR50B2aPOgPro7qWPAAAAAICM0KgBAAAAQMnQqAEAAABAydCoAQAAAEDJ0KgBAAAAQMl0uusj4nqPGhGMX7vzlbmsf82ON0THnrj6tWD8G0s/FZ3z5BUTgvHh97wUneOz2O4f2Vp3Svjx19vif2f63vXhXZ8nLLm7LjllodeAAfHBHcNbMD923MCq19n5/BeiY62PLqz69oA87D3+yarn3PjqoGB89M3R91lWa9WrAMV5fr+R0THfdW0w/u9D/xqd09vCb4NxzysTo3P6vOLRMdSOM2oAAAAAUDI0agAAAABQMjRqAAAAAFAyNGoAAAAAUDI0agAAAABQMuz6WIPWZ1cE42/77Veic956wKJgfOaEm+qS02bbN/UJxi+fGN8pUqeGww9s6B2d8t0jjwnG7e6H4+sA7TSNHRMd+/L4fwTjrd4WndO8zmrOKSvWFP61u+by0dE59+x5Wd3W/937wrvVStI5v/hMMD7s/Lvqtj4Q09HvgWmjroqMNEfnfP3qo4LxHR4r7+6vQDW2/sM90bFem94ZjJ88NPx7XpIOG/lAMH77yviuj5u2Ku/xtifgjBoAAAAAlAyNGgAAAACUDI0aAAAAAJQMjRoAAAAAlAyNGgAAAACUDI0aAAAAAJQM2/PXwDdsCMbHfzO+9e/6/v2D8U8N/VTV6z/37vh23uuHhnvwPu9bE51zx56/D8Yn922Nzjl6Rni7/99OGhudA7S35Ni3RMeOHHhdMP5c2/ronFG3h2szL007jIuOrTo3/NYZ9+z5x+ic/30t/Kv6hNlHROf8bZ/pwfhRA8NvKyJJZxy4NjxwfnQKUDeL/y3+e+BtfcLb8Pe2+N+bR94dP3YBPV3LFeGt+zeseUd0zs93C2/d37Teo3N6x98pB3XAGTUAAAAAKBkaNQAAAAAoGRo1AAAAACgZGjUAAAAAKBkaNQAAAAAomU53fTSziyV9VNIqd39rGjtd0nGSVqffdqq735RVkj1J26uvhuNPhuMdGfTk0/Gx2MDZ8dvb56QTg/Grv/qT6JwjBz4XjH935kejc8Yf8XA8iQbW0LW2W2S3wQ5869mDo2NNf51VSzZd1jR++2D8YzfcH51z3KBw3e7012OjcyZ9/+VgfMzCR6Jz/vO+9wfjvxx9Z3TO66vDu9L2JA1dZyX3tcP+HB1rU3jXue+t3i06p+WOxcE4e0Fmjzorr46Oj8P/Wv3tbfjw3sF47+HDonNaV66qfqEG1ZUzajMkHRKIn+3ue6T/KDSgdjNErQFZmyHqDMjaDFFnQM06bdTc/XZJz+eQC9DQqDUge9QZkD3qDKiPWl6jNtXM5pjZxWa2bd0yAtAetQZkjzoDskedAVXobqN2gaQJkvaQtFzSmbFvNLMpZvaAmT2wURu6uRzQsLpUa9QZUBOOaUD2qDOgSt1q1Nx9pbu3unubpAsl7dPB905398nuPrlZfbubJ9CQulpr1BnQfRzTgOxRZ0D1utWomdnIii8/KWlefdIBUIlaA7JHnQHZo86A6nVle/6Zkg6SNMTMlkk6TdJBZraHJJe0VNIXM8wRORnx87uC8Y/s/+XonLkHzAjGr3jn9Oicb+w/JRi3uxp7235qbcvznuseDcY/M3BhdM67/uPkYHzSrNXBuCRpdfg1+atO2D865afDzoqM9InOmXRh+G0S2qIztjzUWfHaDtwzGP/YgF90MGurYPTPlxwYnTFiTfiYhuxRZ42j1+vhI8RL7xkfndPyJ7bn76pOGzV3PyIQviiDXICGRq0B2aPOgOxRZ0B91LLrIwAAAAAgAzRqAAAAAFAyNGoAAAAAUDI0agAAAABQMp1uJgL4wpb44AHh8C594n8DWDcmvHtXB6ughxsys398MLKx4ddH3BKdcsK7TgjGe90xu5q0JEnPHbdfdOykwecG4ytbPTpn65n3hAe2GRSds3TqbsH43C+fF50T291xl9uPic6YsHhJB7cH1Ml31gTDQ3qHjw0dGbistdZsANSgLfJ8b/128eeBg7YbHIy3Phfe4biRcUYNAAAAAEqGRg0AAAAASoZGDQAAAABKhkYNAAAAAEqGRg0AAAAASoZGDQAAAABKhu35ARRu69kro2M3rw9v3X9IBzt5r9g/PGfUHVWlJUlaN9KiY03qHYwfeNuJ0Tk7aVYwPv9Hk6JzFh3a0Tb8YXv9bGowPuE3c6Nz2taurXodoFqX7DgzMhIv6k8v+kgwPuDKe+uQEYDu6vV6WzDuTfFj54a3jw/Gm/7K9vztcUYNAAAAAEqGRg0AAAAASoZGDQAAAABKhkYNAAAAAEqGRg0AAAAASoZdH9Gp73z2iqJTQA+3acnS6NiPTjk6GD/k3Auic/4y9SfB+KFrvx6dM/SCu4PxqYdfH50Tc+17z4+OzV4wJhg/YKv4lpR7zzo2GB/+nfCuk5I0anF4d0d2dkRd9Qo/Bpd+f5/olMG97gvGN3prdM5LP3tLMN5PKzpIDkDWvHd4d8cNg+Jz+jz3ajAe3j+ysXFGDQAAAABKhkYNAAAAAEqGRg0AAAAASoZGDQAAAABKhkYNAAAAAEqm00bNzMaa2d/MbL6ZPWJmJ6bxwWZ2q5k9nn7cNvt0gZ6JOgPyQa0B2aPOgProyvb8mySd7O4PmtlASbPM7FZJX5B0m7ufYWbTJE2TdEp2qSJrT/xx92D8sJZZ0TnNFt6a+fhnDojOabninuoSawzUWUTLDbOD8Ykf+FJ0zqKP/SoYv+XUn0XnPPr1AcH4vn03dpBd+PG/e59+0Rm7Na8KxvebfUx0zvAjlgXjbevWdZAbIqi1Otp00B7B+LwvnNfBrHDdfOnpA6Mz+l0f3tIfpUWdNYgN24ZbifVjN0XntLb0DcbDG/03tk7PqLn7cnd/MP18raT5kkZL+rikS9Nvu1TSJ7JKEujpqDMgH9QakD3qDKiPql6jZmbjJO0p6V5Jw919uZQUpKRh9U4OaETUGZAPag3IHnUGdF+XGzUza5F0laST3P3lKuZNMbMHzOyBjdrQnRyBhkGdAfmg1oDsUWdAbbrUqJlZs5JC+727X52GV5rZyHR8pKTgCy/cfbq7T3b3yc0KX5MKgDoD8kKtAdmjzoDadWXXR5N0kaT57n5WxdB1ko5OPz9a0rX1Tw9oDNQZkA9qDcgedQbUR1d2fTxA0lGS5prZ5u3XTpV0hqQrzOxYSU9J+mw2KaI7eg0cGIw/9oPdonOu3vecYLytg35+/uvhSxKeOGZcdI60oIOxhkWdRfiG8GNs0tQHo3N2f3JqML5+dGvV67/rHfOrnvPLsbdEx3b/84nB+Pir4ztkta17vOocEEWtVan3NoOiY4eeF3+sV+uZfx/fwejcuq2DXFBnEU07jAvG2/rHdwtum1fs86bXPzg5Orbi3R4eaIvfXvOTq4Px+FGwcXXaqLn7HYrvmPm++qYDNCbqDMgHtQZkjzoD6qOqXR8BAAAAANmjUQMAAACAkqFRAwAAAICSoVEDAAAAgJKhUQMAAACAkunK9vyI6L3bpGC89ZHHclnf3hHfan/x13sH4/Pf/YsObjHct1/zyrDojNOuODwYHzfv7g7WAWrnm+Ib+Y750V11W2dlN+Z8Wu+Mju2oe7ufDFCAjW/fITo2ZdBfq76945e9Jxi3ufG3oYhsAA6UUq/+/aNjT31qVDC+fnj8Ud5vzf7B+MCn4nvgW+Tm1g+Jn6N5aafwW9h4//hb2zS3rA/Gt7q3JTpn07JnomN4M86oAQAAAEDJ0KgBAAAAQMnQqAEAAABAydCoAQAAAEDJ0KgBAAAAQMmw62Oq1+47B+MTLnoiOueTg68Mxr912nHROYMuu6eq9SVpwUnhnXNmHvTr6Jw9+8Z3AoqZuXZ0MH7uOZ+Ozhl3Abs7AkBP1rzqlejYnNfDu8Ht3ie887AkLTt2bDDe9tqC6hIDSqrt1VejYyPuD++S+OSX4zsr9pm0IRgf/dHn4jm4BePb9Y3nFpuz+OUh0TnPPDQyGB9z5VPROfE9m9EeZ9QAAAAAoGRo1AAAAACgZGjUAAAAAKBkaNQAAAAAoGRo1AAAAACgZGjUAAAAAKBkGmp7/qU/2C86dvXnzgrGJzbH76KVreHtUte8Pby9qSS986vh2ztuu0uic2I59Oqgz45tzv+uh46MzhnyrfB2ykMfZgt+AGhUrfMfj46dOn6fbtwi2/CjcfX6x0PB+KiWvaNznjlwQDC+eFL8+eboQS8F47NXh9+KSZLWPL1NMD7srvjbbezwu/BzRLbgrw/OqAEAAABAydCoAQAAAEDJ0KgBAAAAQMnQqAEAAABAydCoAQAAAEDJdLrro5mNlfRbSSOUbCY43d3PMbPTJR0naXX6rae6+01ZJVoP474d373wa9+O7whZrQmKrzM/tr7qt35HBmthdCy2UySy15PqDCgzag3IHnVWvb433h8d2+HG6m9vYyQ+WMujcwZXvwwy1pXt+TdJOtndHzSzgZJmmdmt6djZ7v6z7NIDGgZ1BuSDWgOyR50BddBpo+buy6Wk/Xb3tWY2X1L8TRgAVI06A/JBrQHZo86A+qjqNWpmNk7SnpLuTUNTzWyOmV1sZtvWOTegIVFnQD6oNSB71BnQfV1u1MysRdJVkk5y95clXSBpgqQ9lPzV5MzIvClm9oCZPbBRG+qQMtBzUWdAPqg1IHvUGVCbLjVqZtaspNB+7+5XS5K7r3T3Vndvk3ShpH1Cc919urtPdvfJzepbr7yBHoc6A/JBrQHZo86A2nXaqJmZSbpI0nx3P6siPrLi2z4paV790wPtvHabAAAFqUlEQVQaA3UG5INaA7JHnQH10ZVdHw+QdJSkuWY2O42dKukIM9tDkktaKumLmWQINAbqDMgHtQZkjzoD6qAruz7eIckCQ7zvBVAn1BmQD2oNyB51BtRHVbs+AgAAAACyR6MGAAAAACVDowYAAAAAJUOjBgAAAAAlQ6MGAAAAACVDowYAAAAAJUOjBgAAAAAlQ6MGAAAAACVDowYAAAAAJUOjBgAAAAAlQ6MGAAAAACVj7p7fYmarJT2ZfjlE0prcFg8rOoei1y9DDkWvX48ctnf3ofVKplbt6kwq/j4uev0y5FD0+mXIoUfVmVS6Y1rR65chh6LXL0MO9Vi/VLVWsjorQw5Fr1+GHIpevx45dKnOcm3U3rSw2QPuPrmQxUuSQ9HrlyGHotcvSw5ZKvrnK3r9MuRQ9PplyKHo9bNW9M9X9PplyKHo9cuQQ9HrZ60MP1/RORS9fhlyKHr9PHPg0kcAAAAAKBkaNQAAAAAomSIbtekFrr1Z0TkUvb5UfA5Fry+VI4csFf3zFb2+VHwORa8vFZ9D0etnreifr+j1peJzKHp9qfgcil4/a2X4+YrOoej1peJzKHp9KaccCnuNGgAAAAAgjEsfAQAAAKBkCmnUzOwQM3vMzBaZ2bQC1l9qZnPNbLaZPZDTmheb2Sozm1cRG2xmt5rZ4+nHbXNe/3Qzeya9H2ab2YezWj9db6yZ/c3M5pvZI2Z2YhrP5X7oYP1c74e8FF1naQ651lrRddZBDrk9xqizfDVinaVrNvQxreg66yQHai2b9akz8dwx9zpz91z/SeotabGkHST1kfSwpF1zzmGppCE5r/keSXtJmlcR+4mkaenn0yT9OOf1T5f0HzneByMl7ZV+PlDSQkm75nU/dLB+rvdDTvd14XWW5pFrrRVdZx3kkNtjjDrL71+j1lm6ZkMf04qus05yoNayyYE6c5475l1nRZxR20fSIndf4u6vS/qjpI8XkEeu3P12Sc+3C39c0qXp55dK+kTO6+fK3Ze7+4Pp52slzZc0WjndDx2s3xNRZ2/Irc46yCE31FmuGrLOpOJrrdHrrJMceqKGrLVGr7M0h4Y+phXRqI2W9HTF18uU/y8Wl3SLmc0ysyk5r11puLsvl5IHgqRhBeQw1czmpKe3M70krJKZjZO0p6R7VcD90G59qaD7IUNlqDOpHLVWhjqTCniMUWeZo87erAy11nB1FshBotayQJ29geeOiczvhyIaNQvE8t568gB330vShySdYGbvyXn9srhA0gRJe0haLunMPBY1sxZJV0k6yd1fzmPNTtYv5H7IWBnqTKLWNsv9MUad5YI6K5eGq7NIDtRaNqizBM8dc6yzIhq1ZZLGVnw9RtKzeSbg7s+mH1dJukbJKfUirDSzkZKUflyV5+LuvtLdW929TdKFyuF+MLNmJQ/037v71Wk4t/shtH4R90MOCq8zqTS1VmidSfk/xqiz3FBnb9ZQx7Si6yyWA7WWDeoswXPHfOusiEbtfkk7mtl4M+sj6XBJ1+W1uJkNMLOBmz+XdLCkeR3Pysx1ko5OPz9a0rV5Lr75AZ76pDK+H8zMJF0kab67n1UxlMv9EFs/7/shJ4XWmVSqWiu0zqR8H2PUWa6oszdrmGNa0XXWUQ7UWv1RZ2/gueP/xfO5H6rdfaQe/yR9WMmuKYslfSvntXdQslvQw5IeyWt9STOVnBrdqOQvQ8dK2k7SbZIeTz8Oznn930maK2mOkgf8yIzvg3cpuVRhjqTZ6b8P53U/dLB+rvdDXv+KrLN0/dxrreg66yCH3B5j1Fm+/xqxztJ1G/qYVnSddZIDtVb/takznjsWUmeWJgEAAAAAKIlC3vAaAAAAABBHowYAAAAAJUOjBgAAAAAlQ6MGAAAAACVDowYAAAAAJUOjBgAAAAAlQ6MGAAAAACVDowYAAAAAJfP/AQ15HpQ3bv2dAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1080x1080 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "k=np.random.randint(size=(1,),low=0,high=7900)[0]\n",
    "out0=Sess.run(Lt_vec,feed_dict={X:data.train.images[k:k+1]})\n",
    "out1=Sess.run(Lt_vec,feed_dict={X:data.train.images[99+k:k+100]})\n",
    "out2=Sess.run(Lt_vec,feed_dict={X:data.train.images[k+14:k+15]})\n",
    "diff=out0-out1\n",
    "new_vec=(out2+diff)\n",
    "\n",
    "\n",
    "ans=Sess.run(output,feed_dict={Y:new_vec})\n",
    "\n",
    "fig=plt.figure(figsize=(15,15))\n",
    "ax1=fig.add_subplot(1,4,1)\n",
    "ax2=fig.add_subplot(1,4,2)\n",
    "ax3=fig.add_subplot(1,4,3)\n",
    "ax4=fig.add_subplot(1,4,4)\n",
    "\n",
    "ax1.imshow(data.train.images[k:k+1].reshape(28,28))\n",
    "ax2.imshow(data.train.images[k+99:k+100].reshape(28,28))\n",
    "ax3.imshow(data.train.images[k+14:k+15].reshape(28,28))\n",
    "ax4.imshow(ans.reshape(28,28))\n",
    "\n",
    "ax1.set_title(\"Img_1\")\n",
    "ax2.set_title(\"Img_2\")\n",
    "ax3.set_title(\"Input\")\n",
    "ax4.set_title(\"Generated_image\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Saving\n",
    "saver=tf.train.Saver()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'mymodel/vae'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saver.save(Sess,\"mymodel/vae\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}