FINISHED 08 Your First GAN.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Your First GAN\n",
    "\n",
    "We will be using only a lightly modified version of a fantastic notebook I found online by Wouter Bulten, found here: https://github.com/wouterbulten/deeplearning-resources/tree/master/notebooks. He also has a great blog post on his personal site that goes into deeper detail than we will cover in our workshop, but be sure to check it out too! https://www.wouterbulten.nl/blog/tech/getting-started-with-generative-adversarial-networks/\n",
    "\n",
    "This notebook is very resource intensive. I will be using my Deep Learning Virtual Machine instance from my Azure account, running on a Titan GPU powered VM. We will train for around half an hour, which will give us a good visual validation on what the Generator is actually doing.\n",
    "\n",
    "We start by setting up keras and importing the layers we'll be using for Generative Adversarial Network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\admin\\Envs\\deepvision\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\base.py:198: retry (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use the retry module or similar alternatives.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation, Flatten, Reshape\n",
    "from keras.layers import Conv2D, UpSampling2D\n",
    "from keras.layers import LeakyReLU, Dropout\n",
    "from keras.layers import BatchNormalization\n",
    "from keras.optimizers import Adam, SGD, RMSprop\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import clear_output, Image\n",
    "\n",
    "from tensorflow.examples.tutorials.mnist import input_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def discriminator():\n",
    "    net = Sequential()\n",
    "    input_shape = (28, 28, 1)\n",
    "    dropout_prob = 0.4\n",
    "    net.add( Conv2D(64, 5, strides=2, input_shape=input_shape, padding='same'))\n",
    "    net.add(LeakyReLU())\n",
    "    \n",
    "    net.add(Conv2D(128, 5, strides=2, padding='same'))\n",
    "    net.add(LeakyReLU())\n",
    "    net.add(Dropout(dropout_prob))\n",
    "    \n",
    "    net.add(Conv2D(256, 5, strides=2, padding='same'))\n",
    "    net.add(LeakyReLU())\n",
    "    net.add(Dropout(dropout_prob))\n",
    "    \n",
    "    net.add(Conv2D(512, 5, strides=1, padding='same'))\n",
    "    net.add(LeakyReLU())\n",
    "    net.add(Dropout(dropout_prob))\n",
    "    \n",
    "    net.add(Flatten())\n",
    "    net.add(Dense(1))\n",
    "    net.add(Activation('sigmoid'))\n",
    "    \n",
    "    return net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 14, 14, 64)        1664      \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_1 (LeakyReLU)    (None, 14, 14, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 7, 7, 128)         204928    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_2 (LeakyReLU)    (None, 7, 7, 128)         0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 7, 7, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 4, 4, 256)         819456    \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_3 (LeakyReLU)    (None, 4, 4, 256)         0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 4, 4, 256)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 4, 4, 512)         3277312   \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_4 (LeakyReLU)    (None, 4, 4, 512)         0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 4, 4, 512)         0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 8192)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 8193      \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 1)                 0         \n",
      "=================================================================\n",
      "Total params: 4,311,553\n",
      "Trainable params: 4,311,553\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "net_discriminator = discriminator()\n",
    "net_discriminator.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generator():\n",
    "    \n",
    "    net = Sequential()\n",
    "    dropout_prob = 0.4\n",
    "    \n",
    "    net.add(Dense(7*7*256, input_dim=100))\n",
    "    net.add(BatchNormalization(momentum=0.9))\n",
    "    net.add(LeakyReLU())\n",
    "    net.add(Reshape((7,7,256)))\n",
    "    net.add(Dropout(dropout_prob))\n",
    "    \n",
    "    net.add(UpSampling2D())\n",
    "    net.add(Conv2D(128, 5, padding='same'))\n",
    "    net.add(BatchNormalization(momentum=0.9))\n",
    "    net.add(LeakyReLU())\n",
    "    \n",
    "    net.add(UpSampling2D())\n",
    "    net.add(Conv2D(64, 5, padding='same'))\n",
    "    net.add(BatchNormalization(momentum=0.9))\n",
    "    net.add(LeakyReLU())\n",
    "    \n",
    "    net.add(Conv2D(32, 5, padding='same'))\n",
    "    net.add(BatchNormalization(momentum=0.9))\n",
    "    net.add(LeakyReLU())\n",
    "    \n",
    "    net.add(Conv2D(1, 5, padding='same'))\n",
    "    net.add(Activation('sigmoid'))\n",
    "    \n",
    "    return net\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_2 (Dense)              (None, 12544)             1266944   \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 12544)             50176     \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_5 (LeakyReLU)    (None, 12544)             0         \n",
      "_________________________________________________________________\n",
      "reshape_1 (Reshape)          (None, 7, 7, 256)         0         \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 7, 7, 256)         0         \n",
      "_________________________________________________________________\n",
      "up_sampling2d_1 (UpSampling2 (None, 14, 14, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 14, 14, 128)       819328    \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 14, 14, 128)       512       \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_6 (LeakyReLU)    (None, 14, 14, 128)       0         \n",
      "_________________________________________________________________\n",
      "up_sampling2d_2 (UpSampling2 (None, 28, 28, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 28, 28, 64)        204864    \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 28, 28, 64)        256       \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_7 (LeakyReLU)    (None, 28, 28, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 28, 28, 32)        51232     \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 28, 28, 32)        128       \n",
      "_________________________________________________________________\n",
      "leaky_re_lu_8 (LeakyReLU)    (None, 28, 28, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 28, 28, 1)         801       \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 28, 28, 1)         0         \n",
      "=================================================================\n",
      "Total params: 2,394,241\n",
      "Trainable params: 2,368,705\n",
      "Non-trainable params: 25,536\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "net_generator = generator()\n",
    "net_generator.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "sequential_1 (Sequential)    (None, 1)                 4311553   \n",
      "=================================================================\n",
      "Total params: 4,311,553\n",
      "Trainable params: 4,311,553\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "optim_discriminator = RMSprop(lr=0.0008, clipvalue=1.0, decay=1e-10)\n",
    "model_discriminator = Sequential()\n",
    "model_discriminator.add(net_discriminator)\n",
    "model_discriminator.compile(loss='binary_crossentropy', optimizer=optim_discriminator, metrics=['accuracy'])\n",
    "model_discriminator.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "sequential_2 (Sequential)    (None, 28, 28, 1)         2394241   \n",
      "_________________________________________________________________\n",
      "sequential_1 (Sequential)    (None, 1)                 4311553   \n",
      "=================================================================\n",
      "Total params: 6,705,794\n",
      "Trainable params: 2,368,705\n",
      "Non-trainable params: 4,337,089\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "optim_adversarial = Adam(lr=0.0004, clipvalue=1.0, decay=1e-10)\n",
    "model_adversarial = Sequential()\n",
    "model_adversarial.add(net_generator)\n",
    "\n",
    "for layer in net_discriminator.layers:\n",
    "    layer.trainable = False\n",
    "\n",
    "model_adversarial.add(net_discriminator)\n",
    "model_adversarial.compile(loss='binary_crossentropy', optimizer=optim_adversarial, metrics=['accuracy'])\n",
    "model_adversarial.summary()\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From <ipython-input-9-25e2bcf00b7a>: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\\admin\\Envs\\deepvision\\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\\admin\\Envs\\deepvision\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\base.py:219: retry.<locals>.wrap.<locals>.wrapped_fn (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use urllib or similar directly.\n",
      "Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.\n",
      "WARNING:tensorflow:From C:\\Users\\admin\\Envs\\deepvision\\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 mnist\\train-images-idx3-ubyte.gz\n",
      "Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.\n",
      "WARNING:tensorflow:From C:\\Users\\admin\\Envs\\deepvision\\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 mnist\\train-labels-idx1-ubyte.gz\n",
      "WARNING:tensorflow:From C:\\Users\\admin\\Envs\\deepvision\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:110: dense_to_one_hot (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.one_hot on tensors.\n",
      "Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.\n",
      "Extracting mnist\\t10k-images-idx3-ubyte.gz\n",
      "Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.\n",
      "Extracting mnist\\t10k-labels-idx1-ubyte.gz\n",
      "WARNING:tensorflow:From C:\\Users\\admin\\Envs\\deepvision\\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": [
    "x_train = input_data.read_data_sets('mnist', one_hot=True).train.images\n",
    "x_train = x_train.reshape(-1, 28,  28, 1).astype(np.float32)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(55000, 28, 28, 1)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\admin\\Envs\\deepvision\\lib\\site-packages\\keras\\engine\\training.py:973: UserWarning: Discrepancy between trainable weights and collected trainable weights, did you set `model.trainable` without calling `model.compile` after ?\n",
      "  'Discrepancy between trainable weights and collected trainable'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Variable *= will be deprecated. Use variable.assign_mul if you want assignment to the variable value or 'x = x * y' if you want a new python Tensor object.\n",
      "0: [Dis. loss: 0.6938, acc: 0.5039] [Gen. loss: 2.6236, acc: 0.0000]\n",
      "50: [Dis. loss: 7.9712, acc: 0.5000] [Gen. loss: 0.0000, acc: 1.0000]\n",
      "100: [Dis. loss: 7.9712, acc: 0.5000] [Gen. loss: 0.0000, acc: 1.0000]\n",
      "150: [Dis. loss: 0.7526, acc: 0.5449] [Gen. loss: 1.3987, acc: 0.0352]\n",
      "200: [Dis. loss: 0.5031, acc: 0.7930] [Gen. loss: 1.1550, acc: 0.1953]\n",
      "250: [Dis. loss: 0.4175, acc: 0.8281] [Gen. loss: 1.3228, acc: 0.1875]\n",
      "300: [Dis. loss: 0.2768, acc: 0.9180] [Gen. loss: 2.1077, acc: 0.0391]\n",
      "350: [Dis. loss: 0.3272, acc: 0.8848] [Gen. loss: 1.8549, acc: 0.0664]\n",
      "400: [Dis. loss: 0.3961, acc: 0.8281] [Gen. loss: 1.5457, acc: 0.1875]\n",
      "450: [Dis. loss: 0.5358, acc: 0.7227] [Gen. loss: 1.7701, acc: 0.0703]\n",
      "500: [Dis. loss: 0.4802, acc: 0.7559] [Gen. loss: 0.9739, acc: 0.4570]\n",
      "550: [Dis. loss: 0.4564, acc: 0.7949] [Gen. loss: 1.2879, acc: 0.2070]\n",
      "600: [Dis. loss: 0.4653, acc: 0.7871] [Gen. loss: 1.0820, acc: 0.3555]\n",
      "650: [Dis. loss: 0.3912, acc: 0.8340] [Gen. loss: 1.1770, acc: 0.3164]\n",
      "700: [Dis. loss: 0.5228, acc: 0.7227] [Gen. loss: 1.3944, acc: 0.1602]\n",
      "750: [Dis. loss: 0.4527, acc: 0.7773] [Gen. loss: 1.6857, acc: 0.1250]\n",
      "800: [Dis. loss: 0.5235, acc: 0.7383] [Gen. loss: 1.7046, acc: 0.0898]\n",
      "850: [Dis. loss: 0.5123, acc: 0.7422] [Gen. loss: 1.5754, acc: 0.1250]\n",
      "900: [Dis. loss: 0.4970, acc: 0.7637] [Gen. loss: 2.0302, acc: 0.0898]\n",
      "950: [Dis. loss: 0.5808, acc: 0.7012] [Gen. loss: 1.0012, acc: 0.4180]\n",
      "1000: [Dis. loss: 0.4664, acc: 0.7715] [Gen. loss: 1.2225, acc: 0.2812]\n",
      "1050: [Dis. loss: 0.5530, acc: 0.7227] [Gen. loss: 1.1478, acc: 0.2852]\n",
      "1100: [Dis. loss: 0.5136, acc: 0.7559] [Gen. loss: 0.9187, acc: 0.4492]\n",
      "1150: [Dis. loss: 0.5164, acc: 0.7344] [Gen. loss: 1.2118, acc: 0.3281]\n",
      "1200: [Dis. loss: 0.5401, acc: 0.7480] [Gen. loss: 1.2772, acc: 0.3398]\n",
      "1250: [Dis. loss: 0.4721, acc: 0.7676] [Gen. loss: 1.1356, acc: 0.2812]\n",
      "1300: [Dis. loss: 0.4187, acc: 0.7969] [Gen. loss: 1.3760, acc: 0.2109]\n",
      "1350: [Dis. loss: 0.4783, acc: 0.7930] [Gen. loss: 1.5830, acc: 0.2031]\n",
      "1400: [Dis. loss: 0.5363, acc: 0.7344] [Gen. loss: 1.7378, acc: 0.1055]\n",
      "1450: [Dis. loss: 0.5277, acc: 0.7285] [Gen. loss: 1.0819, acc: 0.3281]\n",
      "1500: [Dis. loss: 0.4266, acc: 0.8223] [Gen. loss: 1.8044, acc: 0.1133]\n",
      "1550: [Dis. loss: 0.5673, acc: 0.7129] [Gen. loss: 1.6262, acc: 0.1523]\n",
      "1600: [Dis. loss: 0.4976, acc: 0.7676] [Gen. loss: 1.1454, acc: 0.3281]\n",
      "1650: [Dis. loss: 0.5886, acc: 0.6875] [Gen. loss: 1.3769, acc: 0.1797]\n",
      "1700: [Dis. loss: 0.5297, acc: 0.7422] [Gen. loss: 1.0606, acc: 0.3789]\n",
      "1750: [Dis. loss: 0.4581, acc: 0.7754] [Gen. loss: 1.8206, acc: 0.1133]\n",
      "1800: [Dis. loss: 0.4605, acc: 0.7695] [Gen. loss: 1.0268, acc: 0.3750]\n",
      "1850: [Dis. loss: 0.4734, acc: 0.7559] [Gen. loss: 1.1940, acc: 0.3242]\n",
      "1900: [Dis. loss: 0.5659, acc: 0.6992] [Gen. loss: 0.7518, acc: 0.5234]\n",
      "1950: [Dis. loss: 0.5479, acc: 0.7578] [Gen. loss: 1.4305, acc: 0.1719]\n",
      "2000: [Dis. loss: 0.5225, acc: 0.7305] [Gen. loss: 1.6446, acc: 0.1602]\n"
     ]
    }
   ],
   "source": [
    "batch_size = 256\n",
    "\n",
    "vis_noise = np.random.uniform(-1.0, 1.0, size=[16, 100])\n",
    "\n",
    "loss_adv = []\n",
    "loss_dis = []\n",
    "acc_adv = []\n",
    "acc_dis = []\n",
    "plot_iteration = []\n",
    "\n",
    "for i in range(2001):\n",
    "    \n",
    "    # Select a random set of training images from the mnist dataset\n",
    "    images_train = x_train[np.random.randint(0, x_train.shape[0], size=batch_size), :, :, :]\n",
    "    # Generate a random noise vector\n",
    "    noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100])\n",
    "    # Use the generator to create fake images from the noise vector\n",
    "    images_fake = net_generator.predict(noise)\n",
    "    \n",
    "    # Create a dataset with fake and real images\n",
    "    x = np.concatenate((images_train, images_fake))\n",
    "    y = np.ones([2*batch_size, 1])\n",
    "    y[batch_size:, :] = 0 \n",
    "\n",
    "    # Train discriminator for one batch\n",
    "    d_stats = model_discriminator.train_on_batch(x, y)\n",
    "    \n",
    "    # Train the generator\n",
    "    # The input of th adversarial model is a list of noise vectors. The generator is 'good' if the discriminator classifies\n",
    "    # all the generated images as real. Therefore, the desired output is a list of all ones.\n",
    "    y = np.ones([batch_size, 1])\n",
    "    noise = np.random.uniform(-1.0, 1.0, size=[batch_size, 100])\n",
    "    a_stats = model_adversarial.train_on_batch(noise, y)\n",
    "        \n",
    "    if i % 50 == 0:\n",
    "        print(\"{}: [Dis. loss: {:.4f}, acc: {:.4f}] [Gen. loss: {:.4f}, acc: {:.4f}]\".format(i, d_stats[0], d_stats[1], a_stats[0], a_stats[1]))\n",
    "\n",
    "    if i % 100 == 0:\n",
    "        # Visualize the performance of the generator by producing images from the test vector\n",
    "        images = net_generator.predict(vis_noise)\n",
    "        plt.figure(figsize=(10,10))\n",
    "        for im in range(images.shape[0]):\n",
    "            plt.subplot(4, 4, im+1)\n",
    "            image = images[im, :, :, :]\n",
    "            image = np.reshape(image, [28, 28])            \n",
    "            plt.imshow(image, cmap='gray')\n",
    "            plt.axis('off')\n",
    "        \n",
    "        plt.tight_layout()\n",
    "        plt.savefig(r'images/gan_output/{}.png'.format(i))\n",
    "        plt.close('all')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<img src=\"images/gan_output/learning2.gif\"/>"
      ],
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import imageio\n",
    "\n",
    "filenames=[r'images/gan_output/{}.png'.format(i*100) for i in range(21)]\n",
    "images2 = []\n",
    "for filename in filenames:\n",
    "    images2.append(imageio.imread(filename))\n",
    "\n",
    "images2.append(imageio.imread('images/gan_output/2000.png'))\n",
    "images2.append(imageio.imread('images/gan_output/2000.png'))\n",
    "images2.append(imageio.imread('images/gan_output/2000.png'))\n",
    "        \n",
    "imageio.mimsave(r'images/gan_output/learning2.gif', images2, duration=0.3)\n",
    "\n",
    "Image(url='images/gan_output/learning2.gif')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x432 with 40 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(15,6))\n",
    "noise = np.random.uniform(-1.0, 1.0, size=[40, 100])\n",
    "images = net_generator.predict(noise)\n",
    "\n",
    "for i in range(40):\n",
    "    \n",
    "    image = images[i, :, :, :]\n",
    "    image = np.reshape(image, [28, 28])\n",
    "          \n",
    "    plt.subplot(4, 10, i+1)\n",
    "    plt.imshow(image, cmap='gray')\n",
    "    plt.axis('off')\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  }
 ],
 "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
}