noise_removal_CAE.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "XLxsZPEJUAvz"
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "from tensorflow.contrib.layers import fully_connected"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "fb6O6KgFUOXj"
   },
   "source": [
    "###Reading the Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 86
    },
    "colab_type": "code",
    "id": "Id9mbx4xUZwT",
    "outputId": "f9d013db-193f-430a-c6dc-bd8f02445df6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting /MNIST_data/train-images-idx3-ubyte.gz\n",
      "Extracting /MNIST_data/train-labels-idx1-ubyte.gz\n",
      "Extracting /MNIST_data/t10k-images-idx3-ubyte.gz\n",
      "Extracting /MNIST_data/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "mnist = input_data.read_data_sets(\"/MNIST_data/\", one_hot=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Y4pYnOmmUkY4"
   },
   "source": [
    "###Initializing kernels and biases"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "3t0yDVgiUqdP"
   },
   "outputs": [],
   "source": [
    "initializer = tf.contrib.layers.variance_scaling_initializer()\n",
    "#kernels\n",
    "ew1 = tf.Variable(initializer([3,3,1,32]), dtype = tf.float32)\n",
    "ew2 = tf.Variable(initializer([3,3,32,64]), dtype = tf.float32)\n",
    "ew3 = tf.Variable(initializer([3,3,64,128]), dtype = tf.float32)\n",
    "\n",
    "dw1 = tf.Variable(initializer([3,3,128,64]), dtype = tf.float32)\n",
    "dw2 = tf.Variable(initializer([3,3,64,32]), dtype = tf.float32)\n",
    "dw3 = tf.Variable(initializer([3,3,32,16]), dtype = tf.float32)\n",
    "\n",
    "dw = tf.Variable(initializer([3,3,16,1]), dtype = tf.float32)\n",
    "\n",
    "#biases\n",
    "eb1 = tf.Variable(tf.zeros(32))\n",
    "eb2 = tf.Variable(tf.zeros(64))\n",
    "eb3 = tf.Variable(tf.zeros(128))\n",
    "\n",
    "db1 = tf.Variable(tf.zeros(64))\n",
    "db2 = tf.Variable(tf.zeros(32))\n",
    "db3 = tf.Variable(tf.zeros(16))\n",
    "\n",
    "db = tf.Variable(tf.zeros(1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Zmh-sBajU1ze"
   },
   "source": [
    "###Defining the Model\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Tc8I3ZO3U5vI"
   },
   "outputs": [],
   "source": [
    "x = tf.placeholder(tf.float32, shape = [None, 784])\n",
    "X = tf.reshape(x, [-1, 28, 28, 1])\n",
    "y = tf.placeholder(tf.float32, shape = [None, 784])\n",
    "Y = tf.reshape(y, [-1, 28, 28, 1])\n",
    "\n",
    "strides = 1 #convolution\n",
    "k = 2 #pooling\n",
    "\n",
    "#model\n",
    "\n",
    "conv1 = tf.nn.conv2d(X, ew1, strides = [1, strides, strides, 1], padding = 'SAME')\n",
    "conv1 = tf.nn.bias_add(conv1, eb1)\n",
    "conv1 = tf.nn.leaky_relu(conv1)\n",
    "conv1 = tf.nn.max_pool(conv1, ksize = [1, k, k, 1], strides = [1, k, k, 1], padding= 'SAME')\n",
    "\n",
    "conv2 = tf.nn.conv2d(conv1, ew2, strides = [1, strides, strides, 1], padding = 'SAME')\n",
    "conv2 = tf.nn.bias_add(conv2, eb2)\n",
    "conv2 = tf.nn.leaky_relu(conv2)\n",
    "conv2 = tf.nn.max_pool(conv2, ksize = [1, k, k, 1], strides = [1, k, k, 1], padding= 'SAME')\n",
    "\n",
    "conv3 = tf.nn.conv2d(conv2, ew3, strides = [1, strides, strides, 1], padding = 'SAME')\n",
    "conv3 = tf.nn.bias_add(conv3, eb3)\n",
    "conv3 = tf.nn.leaky_relu(conv3)\n",
    "encoder = tf.nn.max_pool(conv3, ksize = [1, k, k, 1], strides = [1, k, k, 1], padding= 'SAME')\n",
    "#print(encoder.get_shape())\n",
    "\n",
    "upsample1 = tf.image.resize_images(encoder, size = (7,7), method = tf.image.ResizeMethod.NEAREST_NEIGHBOR)\n",
    "conv4 = tf.nn.conv2d(upsample1, dw1, strides = [1, strides, strides, 1], padding = 'SAME')\n",
    "conv4 = tf.nn.bias_add(conv4, db1)\n",
    "conv4 = tf.nn.leaky_relu(conv4)\n",
    "\n",
    "upsample2 = tf.image.resize_images(conv4, size = (14,14), method = tf.image.ResizeMethod.NEAREST_NEIGHBOR)\n",
    "conv5 = tf.nn.conv2d(upsample2, dw2, strides = [1, strides, strides, 1], padding = 'SAME')\n",
    "conv5 = tf.nn.bias_add(conv5, db2)\n",
    "conv5 = tf.nn.leaky_relu(conv5)\n",
    "\n",
    "upsample3 = tf.image.resize_images(conv5, size = (28,28), method = tf.image.ResizeMethod.NEAREST_NEIGHBOR)\n",
    "conv6 = tf.nn.conv2d(upsample3, dw3, strides = [1, strides, strides, 1], padding = 'SAME')\n",
    "conv6 = tf.nn.bias_add(conv6, db3)\n",
    "conv6 = tf.nn.leaky_relu(conv6)\n",
    "\n",
    "decoder = tf.nn.conv2d(conv6, dw, strides = [1, strides, strides, 1], padding = 'SAME')\n",
    "decoder = tf.nn.bias_add(decoder, db)\n",
    "decoder = tf.nn.sigmoid(decoder)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "VUBeHwDQVJWh"
   },
   "source": [
    "###Defining Loss function and Optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "oHDN9TELVNSv"
   },
   "outputs": [],
   "source": [
    "with tf.device('/gpu:0'): \n",
    "    lr = 0.0001\n",
    "    #loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(labels = x, logits = oned))\n",
    "    loss = tf.reduce_mean(tf.square(decoder - Y))\n",
    "    optimizer = tf.train.AdamOptimizer(lr)\n",
    "    train = optimizer.minimize(loss)\n",
    "    init = tf.global_variables_initializer()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "IXDE9yIeVSg0"
   },
   "source": [
    "##Run the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1397
    },
    "colab_type": "code",
    "id": "frjtZR8QVwJt",
    "outputId": "efbc16db-1105-4ce3-ecb7-38a4122eda02"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 1 loss 0.044419851154088974\n",
      "epoch 2 loss 0.03582439944148064\n",
      "epoch 3 loss 0.030512191355228424\n",
      "epoch 4 loss 0.027293534949421883\n",
      "epoch 5 loss 0.027106216177344322\n",
      "epoch 6 loss 0.0233294814825058\n",
      "epoch 7 loss 0.024739479646086693\n",
      "epoch 8 loss 0.023825304582715034\n",
      "epoch 9 loss 0.022367699071764946\n",
      "epoch 10 loss 0.02194182761013508\n",
      "epoch 11 loss 0.020648164674639702\n",
      "epoch 12 loss 0.02050621435046196\n",
      "epoch 13 loss 0.020333871245384216\n",
      "epoch 14 loss 0.01914030872285366\n",
      "epoch 15 loss 0.01867540366947651\n",
      "epoch 16 loss 0.01949993148446083\n",
      "epoch 17 loss 0.019112640991806984\n",
      "epoch 18 loss 0.0189772117882967\n",
      "epoch 19 loss 0.017940590158104897\n",
      "epoch 20 loss 0.01885383203625679\n",
      "epoch 21 loss 0.01815318875014782\n",
      "epoch 22 loss 0.01670677214860916\n",
      "epoch 23 loss 0.01792907528579235\n",
      "epoch 24 loss 0.017059821635484695\n",
      "epoch 25 loss 0.016798248514533043\n",
      "epoch 26 loss 0.01789703406393528\n",
      "epoch 27 loss 0.01858496479690075\n",
      "epoch 28 loss 0.016445524990558624\n",
      "epoch 29 loss 0.016646752133965492\n",
      "epoch 30 loss 0.01710066944360733\n",
      "epoch 31 loss 0.016923723742365837\n",
      "epoch 32 loss 0.016623083502054214\n",
      "epoch 33 loss 0.016221925616264343\n",
      "epoch 34 loss 0.017218343913555145\n",
      "epoch 35 loss 0.016551688313484192\n",
      "epoch 36 loss 0.015251468867063522\n",
      "epoch 37 loss 0.015523452311754227\n",
      "epoch 38 loss 0.01634015142917633\n",
      "epoch 39 loss 0.01602213643491268\n",
      "epoch 40 loss 0.015951870009303093\n",
      "epoch 41 loss 0.015875590965151787\n",
      "epoch 42 loss 0.015481166541576385\n",
      "epoch 43 loss 0.014842796139419079\n",
      "epoch 44 loss 0.015199782326817513\n",
      "epoch 45 loss 0.015392553992569447\n",
      "epoch 46 loss 0.015026893466711044\n",
      "epoch 47 loss 0.014759073965251446\n",
      "epoch 48 loss 0.015593690797686577\n",
      "epoch 49 loss 0.014943702146410942\n",
      "epoch 50 loss 0.01412356086075306\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHwAAAIACAYAAADwuHuvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<Figure size 1440x720 with 30 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "num_epoch = 50\n",
    "batch_size = 150\n",
    "num_test_images = 10\n",
    "\n",
    "#Noise Properties\n",
    "\n",
    "row,col = 28,28\n",
    "s_vs_p = .5\n",
    "amount = .6\n",
    "\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    sess.run(init)\n",
    "    for epoch in range(num_epoch):\n",
    "        num_batches = mnist.train.num_examples // batch_size\n",
    "        \n",
    "        for i in range(num_batches):\n",
    "            X_batch,_ = mnist.train.next_batch(batch_size)\n",
    "            noised_batch = []\n",
    "            \n",
    "            \n",
    "            #Creating the Noised Batch, introducing Salt and Pepper noise\n",
    "            for i in X_batch:\n",
    "                out = np.copy(i)\n",
    "                # Salt mode\n",
    "                num_salt = np.ceil(amount * i.size * s_vs_p)\n",
    "                coords = [np.random.randint(0, j - 1, int(num_salt))\n",
    "                      for j in i.shape]\n",
    "                out[coords] = 1\n",
    "\n",
    "                # Pepper mode\n",
    "                num_pepper = np.ceil(amount* i.size * (1. - s_vs_p))\n",
    "                coords = [np.random.randint(0, j - 1, int(num_pepper))\n",
    "                      for j in i.shape]\n",
    "                out[coords] = 0\n",
    "                noised_batch.append(out)\n",
    "            \n",
    "            \n",
    "            sess.run(train, feed_dict={x:noised_batch, y:X_batch})\n",
    "        \n",
    "        train_loss = loss.eval(feed_dict = {x:noised_batch, y:X_batch})\n",
    "        print(\"epoch {} loss {}\".format(epoch+1, train_loss))\n",
    "    \n",
    "    noise_test = []\n",
    "      \n",
    "    #Creating Noised Test Images\n",
    "    for i in mnist.test.images:\n",
    "        out = np.copy(i)\n",
    "        # Salt mode\n",
    "        num_salt = np.ceil(amount * i.size * s_vs_p)\n",
    "        coords = [np.random.randint(0, j - 1, int(num_salt))\n",
    "              for j in i.shape]\n",
    "        out[coords] = 1\n",
    "\n",
    "        # Pepper mode\n",
    "        num_pepper = np.ceil(amount* i.size * (1. - s_vs_p))\n",
    "        coords = [np.random.randint(0, j - 1, int(num_pepper))\n",
    "              for j in i.shape]\n",
    "        out[coords] = 0\n",
    "        noise_test.append(out)\n",
    "    \n",
    "    \n",
    "    results = decoder.eval(feed_dict={x:noise_test,y:mnist.test.images[:num_test_images]})\n",
    "    f,a=plt.subplots(3,10,figsize=(20,10))\n",
    "    for i in range(num_test_images):\n",
    "        a[0][i].imshow(np.reshape(mnist.test.images[i],(28,28)))\n",
    "        a[1][i].imshow(np.reshape(noise_test[i],(28,28)))\n",
    "        a[2][i].imshow(np.reshape(results[i],(28,28)))"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "name": "noise_removal_CAE.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}