SSDA-Copy1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Introduction"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this project we will create a Stacked Sparse Denoising Autoencoder (SSDA) with the purpose of denoising grayscale images.\n",
    "\n",
    "This SSDA model is made up of two autoencoders. Rather than training the entire model at once, we will train the two autoencoder layers individually, save their weights, and then use those weights to initialize the stacked autoencoder model.\n",
    "\n",
    "A single autoencoder model includes an input layer, an encoding layer and a decoding layer. After training the first autoencoder and saving its weights, we run the original clean and noisy patches through the input and encoding layers of our training model. The output patches of this model become the input patches to the second autoencoder’s training model.\n",
    "\n",
    "We train the fully assembled SSDA model by initializing the two encoding layers to the weights saved from our individually trained autoencoder models. We provide the trained SSDA model with noisy test images and evaluate its performance."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\ProgramData\\Anaconda3\\lib\\site-packages\\h5py\\__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Input, Dense"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras.models import Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import cv2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Paths to training, validating and testing data\n",
    "\n",
    "For this assignment, we will use images from the Berkeley Segmentation Data Set and Benchmarks 500 (BSDS500). The BSDS500 dataset provides three categories of images: train, val, and test, which we will use for training, validating, and testing the models.\n",
    "https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "path_folder = \".\\BSR\"\n",
    "path_train_img = os.path.join(path_folder, \"train\")\n",
    "path_val_img = os.path.join(path_folder, \"val\")\n",
    "path_test_img = os.path.join(path_folder, \"test\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Patches\n",
    "To reduce the dimensionality of the data, we will convert the images to greyscale.\n",
    "\n",
    "For each image in the train, validate, and test datasets, we will create clean and noisy patches using a window size of 8x8 and a step size of 8.\n",
    "\n",
    "Each noisy patch is created by applying random gaussian noise to its equivalent clean patch. Set the seed variable to allow comparison between program executions."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Method for Sliding window"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sliding_window(image, window_shape, step_size):\n",
    "    rows = image.shape[0]\n",
    "    cols = image.shape[1]\n",
    "    for x in range(0, rows, step_size):\n",
    "        for y in range(0, cols, step_size):\n",
    "            yield(x, y, image[x:x+window_shape[0], y:y+window_shape[1]])\n",
    "            "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Method for Creating patches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def create_clean_patches(path_images, window_shape, step_size):\n",
    "    patches = []\n",
    "    print ('Creating CLEAN ' + str(window_shape[0]) + '*' + str(window_shape[1]) + ' patches for images in ' + path_images)\n",
    "    for i, filename in enumerate(os.listdir(path_images)):\n",
    "        image = cv2.imread(path_images + os.path.sep + filename, cv2.IMREAD_GRAYSCALE)\n",
    "        for (x, y, window) in sliding_window(image, window_shape, step_size):\n",
    "            if(window.shape == window_shape):\n",
    "                patches.append(window.flatten())\n",
    "    print ('All Patched Created added to list')\n",
    "    print()\n",
    "    return np.array(patches)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Citation: http://www.magikcode.com/?p=240\n",
    "def add_gaussian_noise(image_in, noise_sigma):\n",
    "    temp_image = np.float64(np.copy(image_in))\n",
    "\n",
    "    h = temp_image.shape[0]\n",
    "    w = temp_image.shape[1]\n",
    "    np.random.seed(123456789)\n",
    "    noise = np.random.randn(h, w) * noise_sigma\n",
    "\n",
    "    noisy_image = np.zeros(temp_image.shape, np.float64)\n",
    "    if len(temp_image.shape) == 2:\n",
    "        noisy_image = temp_image + noise\n",
    "    else:\n",
    "        noisy_image[:,:,0] = temp_image[:,:,0] + noise\n",
    "        noisy_image[:,:,1] = temp_image[:,:,1] + noise\n",
    "        noisy_image[:,:,2] = temp_image[:,:,2] + noise\n",
    "\n",
    "    return noisy_image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def create_noisy_patches(path_images, window_shape, step_size):\n",
    "    patches = []\n",
    "    print ('Creating NOISY ' + str(window_shape[0]) + '*' + str(window_shape[1]) + ' patches for images in ' + path_images)\n",
    "    for i, filename in enumerate(os.listdir(path_images)):\n",
    "        image = cv2.imread(path_images + os.path.sep + filename, cv2.IMREAD_GRAYSCALE)\n",
    "        #add noise to image\n",
    "        image = add_gaussian_noise(image, 25)\n",
    "        for (x, y, window) in sliding_window(image, window_shape, step_size):\n",
    "            if(window.shape == window_shape):\n",
    "                patches.append(window.flatten())\n",
    "    print ('All Patched Created added to list')\n",
    "    print()\n",
    "    return np.array(patches)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Creating Patches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "window_shape = (8,8)\n",
    "step_size = 8"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Clean Patches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating CLEAN 8*8 patches for images in .\\BSR\\train\n",
      "All Patched Created added to list\n",
      "\n"
     ]
    }
   ],
   "source": [
    "clean_patches_train_8 = create_clean_patches(path_train_img, window_shape, step_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating CLEAN 8*8 patches for images in .\\BSR\\val\n",
      "All Patched Created added to list\n",
      "\n"
     ]
    }
   ],
   "source": [
    "clean_patches_val_8 = create_clean_patches(path_val_img, window_shape, step_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating CLEAN 8*8 patches for images in .\\BSR\\test\n",
      "All Patched Created added to list\n",
      "\n"
     ]
    }
   ],
   "source": [
    "clean_patches_test_8 = create_clean_patches(path_test_img, window_shape, step_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Noisy Patches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating NOISY 8*8 patches for images in .\\BSR\\train\n",
      "All Patched Created added to list\n",
      "\n"
     ]
    }
   ],
   "source": [
    "noisy_patches_train_8 = create_noisy_patches(path_train_img, window_shape, step_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating NOISY 8*8 patches for images in .\\BSR\\val\n",
      "All Patched Created added to list\n",
      "\n"
     ]
    }
   ],
   "source": [
    "noisy_patches_val_8 = create_noisy_patches(path_val_img, window_shape, step_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating NOISY 8*8 patches for images in .\\BSR\\test\n",
      "All Patched Created added to list\n",
      "\n"
     ]
    }
   ],
   "source": [
    "noisy_patches_test_8 = create_noisy_patches(path_test_img, window_shape, step_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Checking patches shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(480000, 64)\n",
      "(240000, 64)\n",
      "(480000, 64)\n",
      "(480000, 64)\n",
      "(240000, 64)\n",
      "(480000, 64)\n"
     ]
    }
   ],
   "source": [
    "print (clean_patches_train_8.shape)\n",
    "print (clean_patches_val_8.shape)\n",
    "print (clean_patches_test_8.shape)\n",
    "\n",
    "print (noisy_patches_train_8.shape)\n",
    "print (noisy_patches_val_8.shape)\n",
    "print (noisy_patches_test_8.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Normalizing patches"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "clean_patches_train_8 = clean_patches_train_8.astype('float32') / 255.\n",
    "clean_patches_val_8 = clean_patches_val_8.astype('float32') / 255.\n",
    "clean_patches_test_8 = clean_patches_test_8.astype('float32') / 255."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "noisy_patches_train_8 = noisy_patches_train_8.astype('float32') / 255.\n",
    "noisy_patches_val_8 = noisy_patches_val_8.astype('float32') / 255.\n",
    "noisy_patches_test_8 = noisy_patches_test_8.astype('float32') / 255."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Autoencoder - 1\n",
    "\n",
    "This model takes flattened 8x8 patches, encodes to size 16x16, and decodes to size 8x8."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Dimensions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_dim_1 = 8*8\n",
    "encoding_dim_1 = 16*16\n",
    "decoding_dim_1 = input_dim_1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Single fully connected layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_img_1 = Input(shape=(input_dim_1,))\n",
    "encoded_1 = Dense(encoding_dim_1, activation='relu')(input_img_1)\n",
    "decoded_1 = Dense(decoding_dim_1, activation='sigmoid')(encoded_1)\n",
    "\n",
    "autoencoder_1 = Model(input_img_1, decoded_1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Encoder model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "encoder_1 = Model(input_img_1, encoded_1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Configuring Autoencoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "autoencoder_1.compile(optimizer='adadelta', loss='mean_squared_error')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Training Autoencoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 480000 samples, validate on 240000 samples\n",
      "Epoch 1/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0150 - val_loss: 0.0058\n",
      "Epoch 2/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0045 - val_loss: 0.0043\n",
      "Epoch 3/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0036 - val_loss: 0.0037\n",
      "Epoch 4/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0031 - val_loss: 0.0032\n",
      "Epoch 5/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 6/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0025 - val_loss: 0.0027\n",
      "Epoch 7/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0023 - val_loss: 0.0025\n",
      "Epoch 8/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0021 - val_loss: 0.0023\n",
      "Epoch 9/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0020 - val_loss: 0.0022\n",
      "Epoch 10/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0018 - val_loss: 0.0021\n",
      "Epoch 11/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0018 - val_loss: 0.0020\n",
      "Epoch 12/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0017 - val_loss: 0.0019\n",
      "Epoch 13/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0016 - val_loss: 0.0018\n",
      "Epoch 14/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0015 - val_loss: 0.0017\n",
      "Epoch 15/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0015 - val_loss: 0.0017\n",
      "Epoch 16/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0014 - val_loss: 0.0016\n",
      "Epoch 17/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0014 - val_loss: 0.0016\n",
      "Epoch 18/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0013 - val_loss: 0.0015\n",
      "Epoch 19/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0013 - val_loss: 0.0015\n",
      "Epoch 20/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0012 - val_loss: 0.0014\n",
      "Epoch 21/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0012 - val_loss: 0.0014\n",
      "Epoch 22/50\n",
      "480000/480000 [==============================] - 10s 20us/step - loss: 0.0012 - val_loss: 0.0014\n",
      "Epoch 23/50\n",
      "480000/480000 [==============================] - 9s 20us/step - loss: 0.0011 - val_loss: 0.0013\n",
      "Epoch 24/50\n",
      "480000/480000 [==============================] - 10s 21us/step - loss: 0.0011 - val_loss: 0.0013\n",
      "Epoch 25/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0011 - val_loss: 0.0012\n",
      "Epoch 26/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 0.0010 - val_loss: 0.0012\n",
      "Epoch 27/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 9.9483e-04 - val_loss: 0.0012\n",
      "Epoch 28/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 9.6627e-04 - val_loss: 0.0012\n",
      "Epoch 29/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 9.3863e-04 - val_loss: 0.0011\n",
      "Epoch 30/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 9.1212e-04 - val_loss: 0.0011\n",
      "Epoch 31/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 8.8675e-04 - val_loss: 0.0011\n",
      "Epoch 32/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 8.6259e-04 - val_loss: 0.0011\n",
      "Epoch 33/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 8.3969e-04 - val_loss: 0.0010\n",
      "Epoch 34/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 8.1815e-04 - val_loss: 9.9940e-04\n",
      "Epoch 35/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 7.9792e-04 - val_loss: 9.7690e-04\n",
      "Epoch 36/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 7.7898e-04 - val_loss: 9.5815e-04\n",
      "Epoch 37/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 7.6136e-04 - val_loss: 9.3715e-04\n",
      "Epoch 38/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 7.4486e-04 - val_loss: 9.2034e-04\n",
      "Epoch 39/50\n",
      "480000/480000 [==============================] - 10s 20us/step - loss: 7.2959e-04 - val_loss: 9.0431e-04\n",
      "Epoch 40/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 7.1530e-04 - val_loss: 8.9469e-04\n",
      "Epoch 41/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 7.0202e-04 - val_loss: 8.7725e-04\n",
      "Epoch 42/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.8959e-04 - val_loss: 8.6112e-04\n",
      "Epoch 43/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.7798e-04 - val_loss: 8.4925e-04\n",
      "Epoch 44/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.6711e-04 - val_loss: 8.3650e-04\n",
      "Epoch 45/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.5684e-04 - val_loss: 8.2594e-04\n",
      "Epoch 46/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.4726e-04 - val_loss: 8.1479e-04\n",
      "Epoch 47/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.3816e-04 - val_loss: 8.0640e-04\n",
      "Epoch 48/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.2954e-04 - val_loss: 7.9434e-04\n",
      "Epoch 49/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.2139e-04 - val_loss: 7.8618e-04\n",
      "Epoch 50/50\n",
      "480000/480000 [==============================] - 9s 19us/step - loss: 6.1369e-04 - val_loss: 7.7596e-04\n"
     ]
    }
   ],
   "source": [
    "history_1 = autoencoder_1.fit(\n",
    "    clean_patches_train_8, clean_patches_train_8,\n",
    "    epochs = 50,\n",
    "    batch_size = 256,\n",
    "    shuffle = True,\n",
    "    validation_data = (clean_patches_val_8, clean_patches_val_8))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plotting the loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_loss', 'loss'])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n2qp8BQ8AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a99c7f29b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print (history_1.history.keys())\n",
    "\n",
    "plt.plot(history_1.history['loss'])\n",
    "plt.plot(history_1.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Output of encoding layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(480000, 256)\n",
      "(240000, 256)\n"
     ]
    }
   ],
   "source": [
    "encoded_imgs_1 = encoder_1.predict(clean_patches_train_8)\n",
    "encoded_imgs_2 = encoder_1.predict(clean_patches_val_8)\n",
    "print (encoded_imgs_1.shape)\n",
    "print(encoded_imgs_2.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Autoencoder - 2\n",
    "\n",
    "Model takes flattened 16x16 patches, encodes to size 32x32, and decodes to size 16x16"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Dimensions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_dim_2 = 16*16\n",
    "encoding_dim_2 = 32*32\n",
    "decoding_dim_2 = input_dim_2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Single fully connected layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_img_2 = Input(shape=(input_dim_2,))\n",
    "encoded_2 = Dense(encoding_dim_2, activation='relu')(input_img_2)\n",
    "decoded_2 = Dense(decoding_dim_2, activation='sigmoid')(encoded_2)\n",
    "\n",
    "autoencoder_2 = Model(input_img_2, decoded_2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Encoder model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "encoder_2 = Model(input_img_2, encoded_2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Configuring Autoencoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "autoencoder_2.compile(optimizer='adadelta', loss='mean_squared_error')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Dividing data into training and validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# encoded_imgs_length = encoded_imgs_1.shape[0]\n",
    "# one_third = int(encoded_imgs_length / 3)\n",
    "# train_data_2 = encoded_imgs_1[:2*one_third]\n",
    "# val_data_2 = encoded_imgs_1[2*one_third:]\n",
    "# print(train_data_2.shape)\n",
    "# print(val_data_2.shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Training Autoencoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 480000 samples, validate on 240000 samples\n",
      "Epoch 1/50\n",
      "480000/480000 [==============================] - 73s 152us/step - loss: 0.1014 - val_loss: 0.0993\n",
      "Epoch 2/50\n",
      "480000/480000 [==============================] - 73s 152us/step - loss: 0.0969 - val_loss: 0.0983\n",
      "Epoch 3/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0961 - val_loss: 0.0976\n",
      "Epoch 4/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0955 - val_loss: 0.0972\n",
      "Epoch 5/50\n",
      "480000/480000 [==============================] - 71s 148us/step - loss: 0.0952 - val_loss: 0.0969\n",
      "Epoch 6/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0949 - val_loss: 0.0966\n",
      "Epoch 7/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0946 - val_loss: 0.0964\n",
      "Epoch 8/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0945 - val_loss: 0.0962\n",
      "Epoch 9/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0943 - val_loss: 0.0961\n",
      "Epoch 10/50\n",
      "480000/480000 [==============================] - 70s 147us/step - loss: 0.0942 - val_loss: 0.0960\n",
      "Epoch 11/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0941 - val_loss: 0.0959\n",
      "Epoch 12/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0940 - val_loss: 0.0958\n",
      "Epoch 13/50\n",
      "480000/480000 [==============================] - 72s 149us/step - loss: 0.0939 - val_loss: 0.0957\n",
      "Epoch 14/50\n",
      "480000/480000 [==============================] - 70s 147us/step - loss: 0.0938 - val_loss: 0.0956\n",
      "Epoch 15/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0938 - val_loss: 0.0956\n",
      "Epoch 16/50\n",
      "480000/480000 [==============================] - 71s 149us/step - loss: 0.0937 - val_loss: 0.0955\n",
      "Epoch 17/50\n",
      "480000/480000 [==============================] - 71s 148us/step - loss: 0.0936 - val_loss: 0.0954\n",
      "Epoch 18/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0936 - val_loss: 0.0954\n",
      "Epoch 19/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0936 - val_loss: 0.0954\n",
      "Epoch 20/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0935 - val_loss: 0.0953\n",
      "Epoch 21/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0935 - val_loss: 0.0953\n",
      "Epoch 22/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0934 - val_loss: 0.0952\n",
      "Epoch 23/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0934 - val_loss: 0.0952\n",
      "Epoch 24/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0934 - val_loss: 0.0952\n",
      "Epoch 25/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0934 - val_loss: 0.0952\n",
      "Epoch 26/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0933 - val_loss: 0.0951\n",
      "Epoch 27/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0933 - val_loss: 0.0951\n",
      "Epoch 28/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0933 - val_loss: 0.0951\n",
      "Epoch 29/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0933 - val_loss: 0.0951\n",
      "Epoch 30/50\n",
      "480000/480000 [==============================] - 71s 149us/step - loss: 0.0933 - val_loss: 0.0950\n",
      "Epoch 31/50\n",
      "480000/480000 [==============================] - 70s 147us/step - loss: 0.0932 - val_loss: 0.0950\n",
      "Epoch 32/50\n",
      "480000/480000 [==============================] - 69s 145us/step - loss: 0.0932 - val_loss: 0.0950\n",
      "Epoch 33/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0932 - val_loss: 0.0950\n",
      "Epoch 34/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0932 - val_loss: 0.0950\n",
      "Epoch 35/50\n",
      "480000/480000 [==============================] - 71s 147us/step - loss: 0.0932 - val_loss: 0.0950\n",
      "Epoch 36/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0932 - val_loss: 0.0950\n",
      "Epoch 37/50\n",
      "480000/480000 [==============================] - 69s 145us/step - loss: 0.0932 - val_loss: 0.0949\n",
      "Epoch 38/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 39/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 40/50\n",
      "480000/480000 [==============================] - 69s 144us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 41/50\n",
      "480000/480000 [==============================] - 69s 145us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 42/50\n",
      "480000/480000 [==============================] - 69s 145us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 43/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 44/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 45/50\n",
      "480000/480000 [==============================] - 69s 145us/step - loss: 0.0931 - val_loss: 0.0949\n",
      "Epoch 46/50\n",
      "480000/480000 [==============================] - 69s 144us/step - loss: 0.0931 - val_loss: 0.0948\n",
      "Epoch 47/50\n",
      "480000/480000 [==============================] - 70s 146us/step - loss: 0.0931 - val_loss: 0.0948\n",
      "Epoch 48/50\n",
      "480000/480000 [==============================] - 71s 148us/step - loss: 0.0931 - val_loss: 0.0948\n",
      "Epoch 49/50\n",
      "480000/480000 [==============================] - 69s 145us/step - loss: 0.0930 - val_loss: 0.0948\n",
      "Epoch 50/50\n",
      "480000/480000 [==============================] - 70s 145us/step - loss: 0.0930 - val_loss: 0.0948\n"
     ]
    }
   ],
   "source": [
    "history_2 = autoencoder_2.fit(\n",
    "    encoded_imgs_1, encoded_imgs_1,\n",
    "    epochs = 50,\n",
    "    batch_size = 256,\n",
    "    shuffle = True,\n",
    "    validation_data = (encoded_imgs_2, encoded_imgs_2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plotting the loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_loss', 'loss'])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJQoSERHpEgWJiIh0yf8HoZ7CD9XKcfwAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a99dc74e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print (history_2.history.keys())\n",
    "\n",
    "plt.plot(history_2.history['loss'])\n",
    "plt.plot(history_2.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Less testing (validation) loss could be because we have trained this encoder on the same Since previous encoder the training and testing data given"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Stacked Sparse Denoising Autoencoder\n",
    "\n",
    "The SSDA model now will take patches of size 8x8 as input. The first encoding layer expands an 8x8 patch to 16x16. The second encoding layer expands a 16x16 patch to 32x32. We then introduce the two decoding layers, which decrease the image from 32x32 to 16x16 to 8x8. \n",
    "\n",
    "The purpose of this model is take a noisy 8x8 image as input and return a clean 8x8 image as output.\n",
    "The weights of the two encoding layers will be initialized with the encoding layer weights that were saved from the corresponding single autoencoder models above."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Dimensions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_dim = 8*8\n",
    "encoding_dim1 = 16*16\n",
    "encoding_dim2 = 32*32\n",
    "decoding_dim1 = encoding_dim1\n",
    "decoding_dim2 = input_dim"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Stack of layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_img = Input(shape=(input_dim,))\n",
    "\n",
    "encoded1 = Dense(encoding_dim1, activation='relu')(input_img)\n",
    "encoded2 = Dense(encoding_dim2, activation='relu')(encoded1)\n",
    "\n",
    "decoded1 = Dense(decoding_dim1, activation='sigmoid')(encoded2)\n",
    "decoded2 = Dense(decoding_dim2, activation='sigmoid')(decoded1)\n",
    "\n",
    "autoencoder = Model(input_img, decoded2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Encoder model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "encoder1 = Model(input_img, encoded1)\n",
    "\n",
    "encoded_img1 = Input(shape=(encoding_dim1,))\n",
    "encoded2_layer = autoencoder.layers[-3]\n",
    "encoder2 = Model(encoded_img1, encoded2_layer(encoded_img1))\n",
    "\n",
    "encoder1.set_weights(encoder_1.get_weights())\n",
    "encoder2.set_weights(encoder_2.get_weights())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Configuring Autoencoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "autoencoder.compile(optimizer='adadelta', loss='mean_squared_error')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Training Autoencoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 480000 samples, validate on 240000 samples\n",
      "Epoch 1/50\n",
      "480000/480000 [==============================] - 75s 157us/step - loss: 0.0083 - val_loss: 0.0059\n",
      "Epoch 2/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0049 - val_loss: 0.0049\n",
      "Epoch 3/50\n",
      "480000/480000 [==============================] - 76s 159us/step - loss: 0.0043 - val_loss: 0.0044\n",
      "Epoch 4/50\n",
      "480000/480000 [==============================] - 75s 157us/step - loss: 0.0039 - val_loss: 0.0040\n",
      "Epoch 5/50\n",
      "480000/480000 [==============================] - 76s 157us/step - loss: 0.0037 - val_loss: 0.0038\n",
      "Epoch 6/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0035 - val_loss: 0.0037\n",
      "Epoch 7/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0034 - val_loss: 0.0042\n",
      "Epoch 8/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0033 - val_loss: 0.0035\n",
      "Epoch 9/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0032 - val_loss: 0.0034\n",
      "Epoch 10/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0032 - val_loss: 0.0034\n",
      "Epoch 11/50\n",
      "480000/480000 [==============================] - 76s 157us/step - loss: 0.0031 - val_loss: 0.0034\n",
      "Epoch 12/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0031 - val_loss: 0.0033\n",
      "Epoch 13/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0030 - val_loss: 0.0032\n",
      "Epoch 14/50\n",
      "480000/480000 [==============================] - 80s 167us/step - loss: 0.0030 - val_loss: 0.0032\n",
      "Epoch 15/50\n",
      "480000/480000 [==============================] - 77s 160us/step - loss: 0.0030 - val_loss: 0.0032\n",
      "Epoch 16/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0029 - val_loss: 0.0031\n",
      "Epoch 17/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0029 - val_loss: 0.0031\n",
      "Epoch 18/50\n",
      "480000/480000 [==============================] - 77s 159us/step - loss: 0.0029 - val_loss: 0.0031\n",
      "Epoch 19/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0029 - val_loss: 0.0031\n",
      "Epoch 20/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0029 - val_loss: 0.0030\n",
      "Epoch 21/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 22/50\n",
      "480000/480000 [==============================] - 77s 160us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 23/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 24/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 25/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 26/50\n",
      "480000/480000 [==============================] - 76s 159us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 27/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 28/50\n",
      "480000/480000 [==============================] - 76s 159us/step - loss: 0.0028 - val_loss: 0.0030\n",
      "Epoch 29/50\n",
      "480000/480000 [==============================] - 76s 159us/step - loss: 0.0027 - val_loss: 0.0030\n",
      "Epoch 30/50\n",
      "480000/480000 [==============================] - 77s 161us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 31/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 32/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0027 - val_loss: 0.0030\n",
      "Epoch 33/50\n",
      "480000/480000 [==============================] - 76s 159us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 34/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 35/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 36/50\n",
      "480000/480000 [==============================] - 76s 159us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 37/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 38/50\n",
      "480000/480000 [==============================] - 77s 159us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 39/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0027 - val_loss: 0.0028\n",
      "Epoch 40/50\n",
      "480000/480000 [==============================] - 75s 157us/step - loss: 0.0027 - val_loss: 0.0028\n",
      "Epoch 41/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0027 - val_loss: 0.0029\n",
      "Epoch 42/50\n",
      "480000/480000 [==============================] - 77s 160us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 43/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 44/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 45/50\n",
      "480000/480000 [==============================] - 75s 157us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 46/50\n",
      "480000/480000 [==============================] - 76s 159us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 47/50\n",
      "480000/480000 [==============================] - 76s 157us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 48/50\n",
      "480000/480000 [==============================] - 75s 157us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 49/50\n",
      "480000/480000 [==============================] - 76s 158us/step - loss: 0.0026 - val_loss: 0.0028\n",
      "Epoch 50/50\n",
      "480000/480000 [==============================] - 77s 161us/step - loss: 0.0026 - val_loss: 0.0028\n"
     ]
    }
   ],
   "source": [
    "history = autoencoder.fit(\n",
    "    noisy_patches_train_8, clean_patches_train_8,\n",
    "    epochs = 50,\n",
    "    batch_size = 256,\n",
    "    shuffle = True,\n",
    "    validation_data = (noisy_patches_val_8, clean_patches_val_8))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Plotting the loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_loss', 'loss'])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nfrloUvvDAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a99d930550>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print (history.history.keys())\n",
    "\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Saving weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "autoencoder.save_weights(\"SSDA_weights.h5\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Testing Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "denoised_patches_test_8 = autoencoder.predict(noisy_patches_test_8)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### PSNR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import math"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def calculate_mse(image1, image2):\n",
    "    squared_error = (image1-image2)**2\n",
    "    sum = np.sum(squared_error)\n",
    "    mse = sum / float(squared_error.size)    \n",
    "    return mse"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "MAX = 1 #We normalised the data\n",
    "def calculate_psnr(image1, image2):\n",
    "    return 20*math.log10(MAX) - 10*math.log10(calculate_mse(image1, image2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def list_psnr(image_list_1, image_list_2):\n",
    "    list_psnr= []\n",
    "    for i in range(1000):\n",
    "        psnr = calculate_psnr(image_list_1[i], image_list_2[i])\n",
    "        list_psnr.append(psnr)\n",
    "    return list_psnr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "psnr_noisy_clean = list_psnr(noisy_patches_test_8, clean_patches_test_8)\n",
    "psnr_denoised_clean = list_psnr(denoised_patches_test_8, clean_patches_test_8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def plot_psnr(list1, list2):\n",
    "    plt.plot(list1)\n",
    "    plt.plot(list2)\n",
    "    plt.title('PSNRs')\n",
    "    plt.ylabel('psnr')\n",
    "    plt.xlabel('patches')\n",
    "    plt.legend(['noisy-clean', 'denoised-clean'], loc='upper left')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEWCAYAAAB1xKBvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a99fc7af28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_psnr(psnr_noisy_clean, psnr_denoised_clean)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}