DenoisingAutoencoder/Denoising-Autoencoder-using-Tensorflow.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from tensorflow.keras.layers import (\n",
    "    Conv2D,\n",
    "    MaxPooling2D,\n",
    "    Conv2DTranspose,\n",
    "    Flatten,\n",
    "    Dense\n",
    ")\n",
    "from tensorflow.keras import Sequential, Model\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "\n",
    "import os\n",
    "import random\n",
    "import tensorflow as tf\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's configure all random numbers generators to support determinism and obtain reproducible results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# settings for reproducibility\n",
    "seed = 42\n",
    "random.seed(seed)\n",
    "tf.random.set_seed(seed)\n",
    "np.random.seed(seed)\n",
    "\n",
    "os.environ['TF_DETERMINISTIC_OPS'] = '1'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Size of train images: (28, 28), Number of train images: 60000\n",
      "Size of test images: (28, 28), Number of test images: 10000\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# load mnist dataset\n",
    "(train_imgs, train_labels), (test_imgs, test_labels) = tf.keras.datasets.mnist.load_data()\n",
    "\n",
    "# fit image pixel values from 0 to 1\n",
    "train_imgs, test_imgs = train_imgs / 255.0, test_imgs / 255.0 \n",
    "\n",
    "# check data array shapes:\n",
    "print(\"Size of train images: {}, Number of train images: {}\".format(train_imgs.shape[-2:], train_imgs.shape[0]))\n",
    "print(\"Size of test images: {}, Number of test images: {}\".format(test_imgs.shape[-2:], test_imgs.shape[0]))\n",
    "\n",
    "# plot image example from training images\n",
    "plt.imshow(train_imgs[1], cmap='Greys')\n",
    "plt.show()\n",
    "\n",
    "# plot image example from test images\n",
    "plt.imshow(test_imgs[0], cmap='Greys')\n",
    "plt.show()\n",
    "plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define leaky ReLU function\n",
    "def lrelu(x, alpha=0.1):\n",
    "    return tf.math.maximum(alpha*x, x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# describe encoder pipeline\n",
    "encoder = Sequential([\n",
    "    # convolution\n",
    "    Conv2D(\n",
    "        filters=32,\n",
    "        kernel_size=(3,3),\n",
    "        strides=(1,1),\n",
    "        padding='SAME',\n",
    "        use_bias=True,\n",
    "        activation=lrelu,\n",
    "        name='conv1'\n",
    "    ),\n",
    "    # the input size is 28x28x32\n",
    "    MaxPooling2D(\n",
    "        pool_size=(2,2),\n",
    "        strides=(2,2),\n",
    "        name='pool1'\n",
    "    ),\n",
    "    # the input size is 14x14x32\n",
    "    Conv2D(\n",
    "        filters=32,\n",
    "        kernel_size=(3,3),\n",
    "        strides=(1,1),\n",
    "        padding='SAME',\n",
    "        use_bias=True,\n",
    "        activation=lrelu,\n",
    "        name='conv2'\n",
    "    ),\n",
    "    # the input size is 14x14x32\n",
    "    MaxPooling2D(\n",
    "        pool_size=(2,2),\n",
    "        strides=(2,2),\n",
    "        name='encoding'\n",
    "    )\n",
    "    # the output size is 7x7x32\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# describe decoder pipeline\n",
    "decoder = Sequential([\n",
    "    Conv2D(\n",
    "        filters=32,\n",
    "        kernel_size=(3,3),\n",
    "        strides=(1,1),\n",
    "        name='conv3',\n",
    "        padding='SAME',\n",
    "        use_bias=True,\n",
    "        activation=lrelu\n",
    "    ),\n",
    "    # updampling, the input size is 7x7x32\n",
    "    Conv2DTranspose(\n",
    "        filters=32,\n",
    "        kernel_size=3,\n",
    "        padding='same',\n",
    "        strides=2,\n",
    "        name='upsample1'\n",
    "    ),\n",
    "    # upsampling, the input size is 14x14x32\n",
    "    Conv2DTranspose(\n",
    "        filters=32,\n",
    "        kernel_size=3,\n",
    "        padding='same',\n",
    "        strides=2,\n",
    "        name='upsample2'\n",
    "    ),\n",
    "    # the input size is 28x28x32\n",
    "    Conv2D(\n",
    "        filters=1,\n",
    "        kernel_size=(3,3),\n",
    "        strides=(1,1),\n",
    "        name='logits',\n",
    "        padding='SAME',\n",
    "        use_bias=True\n",
    "    )    \n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# model class definition\n",
    "class EncoderDecoderModel(Model):\n",
    "    def __init__(self, is_sigmoid=False):\n",
    "        super(EncoderDecoderModel, self).__init__()\n",
    "        # assign encoder sequence\n",
    "        self._encoder = encoder\n",
    "        # assign decoder sequence \n",
    "        self._decoder = decoder\n",
    "        self._is_sigmoid = is_sigmoid\n",
    "        \n",
    "    # forward pass\n",
    "    def call(self, x):\n",
    "        x = self._encoder(x)\n",
    "        decoded = self._decoder(x)\n",
    "        if self._is_sigmoid:\n",
    "            decoded = tf.keras.activations.sigmoid(decoded)\n",
    "        return decoded"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prepare input data for the model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# add noise to the images for train and test\n",
    "def distort_image(input_imgs, noise_factor=0.5):\n",
    "    noisy_imgs = input_imgs + noise_factor * np.random.normal(loc=0.0, scale=1.0, size=input_imgs.shape) \n",
    "    noisy_imgs = np.clip(noisy_imgs, 0., 1.)\n",
    "    return noisy_imgs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# prepare training reference images: add new dimension\n",
    "train_imgs_data = train_imgs[..., tf.newaxis]\n",
    "\n",
    "# prepare test reference images: add new dimension\n",
    "test_imgs_data = test_imgs[..., tf.newaxis]\n",
    "\n",
    "# prepare distorted input data for training\n",
    "train_noisy_imgs = distort_image(train_imgs_data)\n",
    "\n",
    "# prepare distorted input data for evaluation\n",
    "test_noisy_imgs = distort_image(test_imgs_data)\n",
    "\n",
    "# plot distorted image example from training images\n",
    "image_id_to_plot = 0\n",
    "plt.imshow(tf.squeeze(train_noisy_imgs[image_id_to_plot]), cmap='Greys')\n",
    "plt.title(\"The number is: {}\".format(train_labels[image_id_to_plot]))\n",
    "plt.show()\n",
    "\n",
    "# plot distorted image example from test images\n",
    "plt.imshow(tf.squeeze(test_noisy_imgs[image_id_to_plot]), cmap='Greys')\n",
    "plt.title(\"The number is: {}\".format(test_labels[image_id_to_plot]))\n",
    "plt.show()\n",
    "plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define custom target function for further minimization\n",
    "def cost_function(labels=None, logits=None, name=None):\n",
    "    loss = tf.nn.sigmoid_cross_entropy_with_logits(labels=labels, logits=logits, name=name)\n",
    "    return tf.reduce_mean(loss) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# funstion for train and val losses visualizations\n",
    "def plot_losses(results):\n",
    "    plt.plot(results.history['loss'], 'bo', label='Training loss')\n",
    "    plt.plot(results.history['val_loss'], 'r', label='Validation loss')\n",
    "    plt.title('Training and validation loss',fontsize=14)\n",
    "    plt.xlabel('Epochs ',fontsize=14)\n",
    "    plt.ylabel('Loss',fontsize=14)\n",
    "    plt.legend()\n",
    "    plt.show()\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The following block describes basic training pipeline:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 60000 samples, validate on 10000 samples\n",
      "Epoch 1/25\n",
      "60000/60000 [==============================] - 8s 130us/sample - loss: 0.5725 - val_loss: 0.4929\n",
      "Epoch 2/25\n",
      "60000/60000 [==============================] - 6s 93us/sample - loss: 0.4393 - val_loss: 0.3462\n",
      "Epoch 3/25\n",
      "60000/60000 [==============================] - 6s 93us/sample - loss: 0.2448 - val_loss: 0.2063\n",
      "Epoch 4/25\n",
      "60000/60000 [==============================] - 6s 92us/sample - loss: 0.1895 - val_loss: 0.1741\n",
      "Epoch 5/25\n",
      "60000/60000 [==============================] - 6s 95us/sample - loss: 0.1671 - val_loss: 0.1590\n",
      "Epoch 6/25\n",
      "60000/60000 [==============================] - 6s 93us/sample - loss: 0.1548 - val_loss: 0.1485\n",
      "Epoch 7/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1457 - val_loss: 0.1407\n",
      "Epoch 8/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1389 - val_loss: 0.1349\n",
      "Epoch 9/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1342 - val_loss: 0.1310\n",
      "Epoch 10/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1308 - val_loss: 0.1281\n",
      "Epoch 11/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1284 - val_loss: 0.1259\n",
      "Epoch 12/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1265 - val_loss: 0.1243\n",
      "Epoch 13/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1250 - val_loss: 0.1229\n",
      "Epoch 14/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1238 - val_loss: 0.1218\n",
      "Epoch 15/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1227 - val_loss: 0.1208\n",
      "Epoch 16/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1217 - val_loss: 0.1199\n",
      "Epoch 17/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1209 - val_loss: 0.1192\n",
      "Epoch 18/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1202 - val_loss: 0.1185\n",
      "Epoch 19/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1195 - val_loss: 0.1179\n",
      "Epoch 20/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1190 - val_loss: 0.1173\n",
      "Epoch 21/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1184 - val_loss: 0.1168\n",
      "Epoch 22/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1179 - val_loss: 0.1163\n",
      "Epoch 23/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1175 - val_loss: 0.1159\n",
      "Epoch 24/25\n",
      "60000/60000 [==============================] - 5s 91us/sample - loss: 0.1170 - val_loss: 0.1155\n",
      "Epoch 25/25\n",
      "60000/60000 [==============================] - 6s 93us/sample - loss: 0.1166 - val_loss: 0.1152\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# instantiate the model\n",
    "encoder_decoder_model = EncoderDecoderModel()\n",
    "\n",
    "# training loop params\n",
    "num_epochs = 25\n",
    "batch_size_to_set = 64\n",
    "\n",
    "# training process params\n",
    "learning_rate = 1e-5\n",
    "# default number of workers for training process\n",
    "num_workers = 2\n",
    "\n",
    "# initialize the training configurations such as optimizer, loss function and accuracy metrics\n",
    "encoder_decoder_model.compile(\n",
    "    optimizer=tf.compat.v1.train.AdamOptimizer(learning_rate=learning_rate),\n",
    "    loss=cost_function,\n",
    "    metrics=None\n",
    ")\n",
    "\n",
    "results = encoder_decoder_model.fit(\n",
    "    train_noisy_imgs,\n",
    "    train_imgs_data,\n",
    "    epochs=num_epochs,\n",
    "    batch_size=batch_size_to_set,\n",
    "    validation_data=(test_noisy_imgs, test_imgs_data),\n",
    "    workers=num_workers,\n",
    "    shuffle=True\n",
    ")\n",
    "\n",
    "# visualize train and val losses\n",
    "plot_losses(results)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's view the decoding process on the same data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# instantiate the model with final sigmoid activation\n",
    "encoder_decoder_model2 = EncoderDecoderModel(is_sigmoid=True)\n",
    "\n",
    "# define the number of images to decode\n",
    "img_num_to_decode = 10\n",
    "\n",
    "# get 10 test reference images\n",
    "test_imgs_data_decode = test_imgs_data[:img_num_to_decode]\n",
    "\n",
    "# get 10 distorted input test data and cast to float32 as tf layer's dtype is float32\n",
    "test_noisy_imgs_decode = tf.cast(test_noisy_imgs[:img_num_to_decode], tf.float32)\n",
    "\n",
    "decoded_images = encoder_decoder_model2(test_noisy_imgs_decode)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Original Images\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x288 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Noisy Images\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x288 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reconstruction of Noisy Images\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABH4AAACACAYAAAB9Yq5jAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO2debRcVZX/9yUQIQaBQCABAklIIAaQUUxIDCCCgooIQtMyxG5sXCoqmtUL0O5e3fzRC2ztJSqDDJrGAfkhCEFthISEMNiQmIGQeWLKCIQhCe0AfX9/5HH8np1XxXv1blXde97ns1YWu945r+q++7373FOXPWR5nhsAAAAAAAAAAKTHDu0+AAAAAAAAAAAAaA48+AEAAAAAAAAASBQe/AAAAAAAAAAAJAoPfgAAAAAAAAAAEoUHPwAAAAAAAAAAicKDHwAAAAAAAACAROnRg58syz6aZdnSLMtWZFl2RVEHBa0FHasPGqYBOlYfNEwDdKw+aJgG6Fh90DAN0LH6ZHmeN/aLWdbHzJaZ2Slm9oKZzTKzv83zfFFxhwfNBh2rDxqmATpWHzRMA3SsPmiYBuhYfdAwDdAxDXbswe8eZ2Yr8jxfZWaWZdkvzOyTZlbzAsiyrLGnTNBj8jzPagx1S0c0bCsv5Xk+sJOf44sVAl9MAnwxAfDFJMAXEwBfTAJ8MQHwxSSo5Ys9SvXaz8yel9cvdPysLn369LE+ffr04GOhq3TxXHdbxx122MF22IHyUK3gbQ07dHy2xjR8seTgi9Xn7XPdcb4L9cUdd9zRdtyxJ/8fBrpKF32m2zqiYetopi9mWWZZVut7D3TG2+esu+eti7/DfbHENNMX0bF0dFtHvme0ji76Yo8ifrpElmWXmNklzf4caB5omAboWH3QMA3QsfqgYRqgY/VBwzRAx+qDhuWnJw9+1pjZEHm9f8fPIvI8v8nMbjLbFvb11ltv9eAjoTt08Vy/o45ew//7v/8r6hDhHShKQzN8sZ3gi9Wni+e6IV988803izhE6AJF6YiG7aOZvtho3cveTKPnrIu/x32xxDTTF9GxdHTbF/me0Tq66i89iaGbZWYjsywblmVZXzM7z8ym9OD9oD2gY/VBwzRAx+qDhmmAjtUHDdMAHasPGqYBOiZAwxE/eZ6/mWXZpWb2OzPrY2Y/yvN8YWFHBi0BHasPGqYBOlYfNEwDdKw+aJgG6Fh90DAN0DENGm7n3tCHUeG7bdSp0t4t0LCt/CHP82OLeCN0bB/4YhLgiwmALyYBvpgA+GIS4IsJgC8mQU1fpAUFVB5fMb5v377B3nnnnWv+3htvvBHsP//5z9EYef4AzcN3ctHOHT5PGV8EACgvup6zXgMAlBf65AEAAAAAAAAAJAoPfgAAAAAAAAAAEoVUL2grPuWjHjvu+NfLdb/99gv2+eefH807/vjjg73LLrsEe+PGjdG8GTNmBHv27NnR2LJly4K9ZcuWYNNesloQgl4e+vXrF+wjjzwyGtP0zAULFkRjmzZtCjYaAgC0Hr9X0zVb0+3/93//N5pX9TVb05A9/m+r+t8K5UWvw5122ika09d/+ctfgk0JC+gMIn4AAAAAAAAAABKFBz8AAAAAAAAAAInCgx8AAAAAAAAAgEShxg+0HM1H1Ro8ZnEdkKFDh0Zj5557brBPOeWUYO+5557RvLfeeivYmu86cODAaN4hhxwS7L/7u7+LxqZOnRrsG2+8Mdhr166t+VnQHrT2wN577x2NaS2oVatWBfu1116L5pH73BwGDBgQ7GuvvTbYH/7wh6N5WsfnG9/4RjT229/+Ntjqz9A8fD2PVvpHvbpv+Gl7UW3q1U9Dp3emXs2cnXfeOdh//OMfo3l/+tOfmntgNXjXu94VvT7ssMOCrXUQly9fHs2r4h6p1nXu6ep13s71FIrD13vS1/WuE60N6rXX31O/Hzx4cDRvxIgRwT700EOjsV133TXY69evD/ZDDz0UzVuxYkWnxwS9CyJ+AAAAAAAAAAAShQc/AAAAAAAAAACJUulULx9aVy/sTsPr1CbcrTWoNhqWOHz48GiepnCdc8450Zimfr355pvBXr16dTRPW7NrWpkPjxw1alSwNcXMf9YzzzwT7J/97GfRPG1dSvhue9B0ottuuy0a03D0q6++OtiavmdGClFRaKiymdnll18e7LPPPjvYPsWzf//+wZ4wYUI0Nn369GBr2gBrd7Godvvvv380puHjmtZRBDvuGG9Ddtttt2B7v9TPRv/mo/dqM7PDDz882Fu3bg32smXLonm+pTdsQ9ueDxs2LBo777zzgq3p6zfffHM0b8mSJcFutg/ovu2II46IxiZOnBhsTSlZuXJlNK+KqV5KvX1dve8gup6+5z3viebpmvfqq68G+4033ojmVf3cpYBq6r8nDBo0KNjvfe97g+19W68FX45AvwMdcMABwdb7oFn8Xcbvs/R60utV1xQzs09/+tPBfumllwyaS730P08rvz8S8QMAAAAAAAAAkCg8+AEAAAAAAAAASJRKpHppuJSGyvouA+9+97trjmlnhM2bNwf7z3/+czSvVriVfq7Z9tXdFU1DIhx9G7W6JPiQxX333TfYXsM1a9YE+0c/+lGw77zzzmiepgNoKOZXvvKVaJ6GZnp99Rj1/Qi9bT8+fHLs2LHBHj9+fDSmnVJUY/yyOHQt9ClCRx11VLB1bfV+pK933333aExDo3We73ZTr3MGdI76xMUXXxzsE088MZp32WWXBbvoVK+99torej1mzJhgP/fcc9HYU089FWx8uDmov91yyy3R2LHHHhvs+++/P9iTJk2K5pHqtQ2/T9TOoldddVU0dtJJJwV7w4YNwf7Vr37VpKN7ZzRF6YILLojGdK3XtTgFv+xqepdPU9VOTJoK99GPfjSap2k7jz76aLBvvfXWaN7ixYuDzd6zdajGqpVP2dP9zd/8zd8E+8gjj4zm6T3Of+fRe3C9PZL6Vb3vozrPdwbzKfZQDHr+VRu/Pug8n8aur5u9fyXiBwAAAAAAAAAgUXjwAwAAAAAAAACQKDz4AQAAAAAAAABIlFLW+PE1PDRnTtuLaus7M7PRo0cH+6CDDorGtOX3ggULgv3iiy9G8zQnT1tqalto//4+p1lzdrUmgdYWMutdtYD073v99deDvXTp0mjelClTgu3rOzzyyCPBnjNnTrDrteL+05/+FGytEWQWa61amMXXyNSpU4Pd1ZpQHr2mqT/SM7T1t5nZP/7jPwbbt9vctGlTsB988MFgp+5vzUZzlbXtqF8n1TfXrl0bbJ8r/9prrwXba/iBD3wg2C+88EKwvT/rWu7r0FAfoXNGjhwZ7K997WvBfuWVV6J5vs1wT9H1UGutmZl95CMfCfbdd99d6OfC9mgNCzOzK664Iti+NonuxUaNGhVs1tPO8TV+dM+q588sPoezZ88O9qJFi6J5zdw/+JoUp512WrB9O3dt2/70008HO7VrwZ9v1VS/j5iZfeYznwn2l770pWD7Omb6nvvtt1+w/Tr7rW99K9i6b+7suGqha2292qS9uUae/86ptSG1duyAAQOiebpX0e8G/v30tf++onsV1V/3RGbxHsbXP9XaPVpv64c//GE0b926dVZG6p0vxV+XjVyn3ge0pqTWLfPz9Dvi8OHDo7F99tkn2Kqn/86v18iKFSuiMd3P6rxm+CIRPwAAAAAAAAAAicKDHwAAAAAAAACARKlEqpeGXGnYnaZ2mcVhycOGDYvGXn755WBrCz4fPjlixIhgH3744cH2bYq1JZ9PI9A2jvPnzw+2hu+amf3P//xPsH//+99HY75VcdXRcDUNdXzppZeiedOnTw/2zJkzozH9va6GE+vnamt3s1g3n/Knrd/9WCP0ttDZItB1QENbv/CFL0TztMWwvy40VUTTCtGje/iwV02DHTt2bLCPP/74aJ62dV6yZEmwfei7hrb6cOpzzz032HpN6JpuZvbYY48F+9e//nU0tnHjxmCnlorQHXwqx5VXXhlsDXn2LaR92HlP0evpuOOOi8YOPfTQYHsd8dvi0VRNszjVzqeBqe/ovVtTqqE2W7duDbauV2ZxWsD1118fbL/OFe0Duib4NtT/8A//EGyfXvLwww8He/369U07vrKha5f/XnDGGWcEW+9j/v6pfqQpJP67hK7JPg1Mf0/Peb20JZ9Grej79wZ/Vk30PJuZDR06tNN5viSE7in1u6SW+DCL9fFjy5cv7/Q9fEqY6upTDPX4dR3R1PjOjr+s6N9ayzaLz6tfdzQtWX1R95Nm8Xc9nadrtVm8l/WlCvSzdS3UUhNmsY/p93+zeM+laWC+vEgREPEDAAAAAAAAAJAoPPgBAAAAAAAAAEgUHvwAAAAAAAAAACRKKWv8+BoMmpeotQaef/75aJ62/9a8VrM4B09b3/nW0FpDSOv41Mst9Ghrt1NPPTXYH/rQh6J5mtupeYZmZrNmzQp2yq2I/XlUrRvNR1Wt9t1332CPGzcumqe5kz/4wQ+iMd9CFRqnXqvGWm0bzeL6EhMmTAj2pZdeGs1Tf/b1mK655ppgVyW/uYz4nPLPf/7zwT755JOD7etAbNiwIdjPPPNMsH27cH1/X+NHW2f69VrR9sj6uWZmDzzwQLB7Q/2CWvh6Oh//+MeDredl8uTJ0byi6yKpz/qW4Xr/9DWJUq8f0ip03dV20mbb199SdC+i9S1YWzvHX69a8+HBBx+MxrSug65fzbjmdZ97zDHHBNvfWwcOHBhs3ZOamU2bNi3YzahDUVa05otv6+xr/ryNPz9ac07rmOl3ArO4VqlfC7UejNYj0domZnFdnz322CMa0zoyqdUV9fg6SwcccECw/XWv3xtmzJgR7IceeiiapzV01H8fffTRaJ7W3fH7j0b82++f9LtvT9+7Hfjj1P2GXve6bpnFtXb8vvHEE08Mtu5ztHavWbyn1Ptid753675X9y+DBw+O5uk1qLUMzeK19l//9V+D3ZYaP1mW/SjLso1Zlj0tPxuQZdmDWZYt7/jvHvXeA9oPOibBUDSsPvhiEuCLCYAvJgG+mAD4YhLgiwmAL6ZNV1K9JpvZR93PrjCzaXmejzSzaR2vodxMNnSsOi8ZGqbAZEPHqoMvpsFkQ8eqgy+mwWRDx6qDL6bBZEPHZHnHVK88z2dmWTbU/fiTZnZih/1fZjbDzC4v8LgiNOxL26HNmzcvmqdt1HxKgIY/akhjvRBlfQ/fvk3bGY8cOTIa+8QnPhFsTT/Q1DGzOKTat5+fO3dusBtpXd7JWNt1bBXa2vCHP/xhsL1Oa9euDfatt94ajZU0RHKLmW1yP6u0hnqefSiuhk+OHj062JomYhaHK/v2z7VCYNtJVXxR18wTTjghGrvkkkuCrWujhjSbxS0wNcTZp+mqpt73Bg0aFGwN9fWtpjU8W9fnzt6zACrji3qeNPXRLA6P1vaiq1evbuoxaXrJwQcfHI1pCoNvZV00VfHFZqIh5mZxmrz3G/Vn3X+1+X5ZGV/U8+Sv7Wam2vh75plnnhnsc845J9hDhgyJ5j39dPgf/nbddddFY+vWrQt2EamgVfFFTT05+uijozFNq9L0qyVLlkTz9FzqXl/biJvFabBeGz3/U6ZMCXa9PY+/xjRlqKCUklL5oqbu+PP3H//xH8EeP358NKb7mMcffzzY+v3TLN7TaGqQb8Wu3zOLWCubvd622xdr/X0+LXnixInBHjt2bDSmaZjqs1u2bInmLVy4MNjz588P9lNPPRXN03ufv5Z0P6OpZH6ersO77bZbNHbKKacE+7vf/W6wX331VSuaRos775Pn+durznoz26feZCgt6Fh90DAN0LH6oGEaoGP1QcM0QMfqg4ZpgI6J0OPiznme51mW1Xz8mGXZJWZ2Sa1xKAf1dETDaoAvpgG+WH3wxTTAF6sPvpgG+GL1wRfTAF+sNo0++NmQZdngPM/XZVk22Mw21pqY5/lNZnaTmVk9h+8qGk7nw+40xNF3C6oViup/rq813KxeZyKfwqXdD6666qpgv/e9743maSio2v7963U+6mHIX5d0LFrDovGdhP7lX/4l2NoNyqNpQT5FpUK0zRe7ir9Ga12z/ucaljx79uxga3coszj957bbbovGfMhtiSmdL2qKkKYDmMWdCxTvR9q9RFOJtBuGWay1f+9PfepTwT7ttNOCrSmdZvH9wHf1KrorVQ1K6Yt633n/+99fc94jjzwS7Gaknej1dMYZZwTbd6FZvHhxsLVTilnLUopK54tFUy8FQsPivd9oVypNX2mRf3WHtvminlufjqrp/WPGjInGdA84ffr0YPu1UlNy6nWe0c/+yEc+Eo398z//c7A1VdffW++4445ga9qXWcs6uZXOF1XfESNGRGOarq6lJzR1w8xs6tSpwVYNjzzyyGjexz72sWD7DkGaEvvSSy8Fe82aNdE8TSn03zP0s5vow23zRe0WeuGFF0ZjmqbnywxodzX9PtfV8+fPZUlLR3SXtviinjufFq7duvx9TPcVui/VLm1mZrfffnuwNe3Sf3fQ4/DdxRYsWBDsr3/968H2JVz03uqvCX1Pfz0WTaPvPsXM3k6um2hm9xZzONBi0LH6oGEaoGP1QcM0QMfqg4ZpgI7VBw3TAB0ToSvt3G83s9+b2SFZlr2QZdnFZna1mZ2SZdlyM/twx2soMeiYBMMMDSsPvpgE+GIC4ItJgC8mAL6YBPhiAuCLadOVrl5/W2Po5IKPBZoIOibB6jzPj+3k52hYIfDFJMAXEwBfTAJ8MQHwxSTAFxMAX0ybHhd3bic+j7JezrHm0zWSb1mvTom2eTOLc6FXrlwZbF+7Qutf+BaMXW3/l0juaLfRHMhjj43vMxdddFGwNafytddei+Zp3nVvPY9lwmugObZa60Pba5qZ7bHHHsH2bcKhcTQ/3tceUL9SnbQ+i5nZLbfcEuwnn3wy2H7NVO21vaZZ3FpcP3fcuHHRvBdffDHYvv5bb8LXhPvsZz8bbF8PTWv5PPTQQ8EuYj30x3HIIYcEW9ds1c3M7Fe/+lWwtaZMUccFsR9pG1mzuDaM31PNmTMn2OqX6PJX9Lrfe++9o7G///u/D/anP/3paExrRWr9Fl03zeLW3TqmrbnNzA466KBgf/Ob34zGtNWxrsXaztjMbObMmcHW1tWwjUGDBkWva9Xw8HtP/e6ia7KvA6rv379//5rHodr4NVOvpRLW4ioc1eADH/hAsE866aRonq5tjz32WDT2/e9/P9ha78evh7pW1qoP+05jUB9dT31dR10zfV0crb+ka5yvX6j7V9XT66TH0a9fv2hM9zb6Pb87tXq0TpdfL4qmuRWEAAAAAAAAAACgbfDgBwAAAAAAAAAgUSqd6uVpV0pUvTQwDbtctGhRNE9bOq5atSoa0zA1QgO3oaF2mt4zadKkaJ62xdMQWE07Mdu+7SWUC73u1R98GKSGUzajfbu+v16DqbXs9Kk5hx12WLAPPPDAaEzPiaYYaIi0WdwiXNsQ18O3KH722WeDPW3atGC/733vi+ZpW2Lf9lNbybeoDXHb2GWXXaLXJ5xwQrD9NaqpkdqStIhr2bc81Xa6++67b7B9C2ltqVqvXTU0jqYgaVtjs1h7DT83M/vBD34Q7GaHo1cVTTUZNWpUNDZhwoRg+9T/vn37BnvPPfcM9m677RbN07VX9zd+D3n66acH27cV1lbHmgqkaWRmaNwZqq/XUNO2dB3W9c6/1jbtvmyBprL4+7Nqr+lIvuV46uldei2bxZqceuqpwfZpl3rd+xR1TT9WvX2q0cCBA4OtaUK+dMfatWuDnfr+o2j0fuT3CkuWLAm2v871utBUL03LMjP7/Oc/H2zdA/kUdN2L+P3lmDFjgr3XXnsF2/tsvRIxv/zlL4O9ZcsWayZE/AAAAAAAAAAAJAoPfgAAAAAAAAAAEiWpVK924St3H3HEEcHWMMHZs2dH8+69995g+9CuqqeNFIEPk9Mwy7POOivY2r3CLA7R+81vfhPsf/u3f4vmpR4C205Uu65ey15vDZ3VbhcjR46M5vnQ5p7ij8O/ThVdq8zMJk6cGGyfPqThyg8//HCwp0+fHs3ranpXPfT60TBdfx1oaO9pp50WjWkYberdaXxIu6bF+vvMjBkzgl1EWof6rIY/m8WpJ9qhxnfZ0HW5t/heK1D/+MQnPhFs7d5nFq+nd9xxRzSmXaRIw+scTffxewxNnfO+qGus+oSmM5jFJQI0DdankLz66qvB1rQWs9ivHnjggWDPmjUrmsceaXs0hVVTserNU38zi78jjBgxItijR4+u+X5+ndTUPk3Z7W2a+X2LpqVriqPviqYplNr9yyzet2i5Du2GZ2Z26KGHBnvz5s3Bvv/++6N5d955Z7D9fZbvevXRtcqvhTfffHOw1Y/MYl/SFErfoVbX6zPPPDPY/v5W736nzwDU//z+V9O7tIuqmdltt90W7GaUrFCI+AEAAAAAAAAASBQe/AAAAAAAAAAAJAoPfgAAAAAAAAAAEoUaPwVwwAEHRK8/97nPBVvzrn2rzPXr1we7t+XldgXfDvi8884L9hlnnBFs33bv5z//ebCvu+66YBddCwb+Sr26OPVaGir12nKec845wdb8eDOzRYsWBbuIVpn++Gr5Zmq52b417UknnRRsX8fs5ZdfDvaNN94Y7Gb4mLY5Pv7444Ot7dvNYj20HbJZ+rVi9O/zue6qib9mV69eHex+/foF29dBUh/Qa0G18Z/9xS9+MRobMmRIsDWH3bc1feONN2oeLzSO1rQ4+eSTg+3XN21trHUHzLbXCrZH6zr4NtHXXnttsI8++uiav6dti+fPnx/N032j3u98DRPV6oUXXojGFi5cGOwbbrgh2FoXyAz/6wzdl/p6LdrKWc///vvvH80bOnRosPVe5Wvp6Vr4yiuvRGNz5swJttYf0TprfixFPf3eRGvyrFixItjHHHNMNE9r3/kaPzpX/dLf7/S1ztPrwCzeo/o6WkXUQUwZXeP8dz2tpeRrPe27777B1pbt559/fjRPNVTf8XtGvU/6ej+6n1Gf1f2Vmdldd90V7MmTJ0djet02+3kAET8AAAAAAAAAAInCgx8AAAAAAAAAgERJKtWrlWGMGrrpQ7YOPvjgYGsL93Xr1kXzikhLSQ0Nu/vgBz8YjX3mM58Jtrbgu++++6J52uLPt0yF1tBIO3cfqql+NH78+GBrCKeZ2dq1a2t+VhFrQorh0Z3h21wOGDAg2D4cWduAz5s3L9hFnCsfYqupfUcddVSwfYi3hts+/vjj0Zi2Wk0RTZPUtdEsTjfxqQSq8UUXXRRsn9KjIdCaBuZbDL///e8Pdr22qZp+pvdIs7jlNS3DG8enzmr6gq6t/hzPnTs32D60Ht4ZXSs1LcvMbOPGjcGeNm1aNFYrJafePU3XQO/3ej/Ve6SZ2X//938Hu+hU6dTw9yNdG++5555oTNc8Tfnw66Smver14vermqK3YcOGaOzZZ58N9siRI4NdL53Xr+sp7G18WrKmSWp5jY997GPRPP0O5/eeSr1UOV1jd95552Brm3czswsvvDDYmiZvZvbcc88Fu57/6XH0pvuinnN/fuqdr+XLlwf7O9/5TrD9XvaTn/xksPWa8Oup4tut63d7Xde///3vR/OWLl1a8z1aCRE/AAAAAAAAAACJwoMfAAAAAAAAAIBE4cEPAAAAAAAAAECiJFXjp5lo6z+zuI6M1jUwi3NONbeaVpnb4/OnteXsWWedFY1pC0zNnX/00Uejea+//nqBR1gM9dqbN7t1Xyuo1wK9q9e5Pw/qc9oe09d20Tz4ZrQTTxm9Fn1LU81f9zUKHn744WDreuev7UbWOF+H5stf/nKwdX3w6Pp6++23R2PtzKduBXqen3zyyWhM8899m3vV/Oyzzw62rrVmsc+tWbMm2FoPprP3V7Tmhbaovvfee6N5+HAx7LrrrtHrc889N9gDBw4MttZUMjP7wx/+EOx6Wug10YzaalVF/3Zfi6Po2hz6Wdpm3Cyuhbds2bJo7MEHHwy2r/sCMf5a1rpXN954YzSm9y79PV9DptZ9zO9d9T3872i9uxNOOCHYEyZMiOZpnZtHHnkkGkuhppPfN+peRVve33333dE83Vf4tVL9VDXx+5v3vOc9wVaNfdt3rVPoaw3pMaov+npPK1eu7HQedI5eF7oHuuaaa6J5TzzxRLAvu+yyYI8ePTqapzV/XnvttWhMa6Zde+21wdZ6U2bl8TcifgAAAAAAAAAAEoUHPwAAAAAAAAAAiUKqVx00tOuCCy6IxrRdnw/J0/B3bfdYxhSkduNDJzXkcsSIEdGYtkvUlny+PZ9PBWoV/nM17UFTKnzY70MPPRTsVK6RRkL964XBa9imtsQ1I2y9J+g1O2rUqJpj/rxqKPPgwYOD7bXR0NZ66X+61p5++unR2Ac/+MFga/qZD5udPn16sH1qQ+qo7/jUnU2bNgXbh6BryPL48eOD7ddlbSWsaX6PP/54NE/TM/17aOt4TcXz4dC9OU2op6jP+vun6tuvX79ga6qsmdns2bOD7Vsl16KIFE/oPjvttFOwx40bF40NHz482D/96U+jMV0j2qnV29dNla4X9QlN+2qUen+7+nP//v2jsUMOOSTYxx13XLA1xc8sXpPnzZsXjb3yyivdO9iKoVr9+Mc/jsZ037LPPvtEY9p2e8GCBcH2KdDnnHNOsA844IBg+9Ry9VOfQqQtxDWdT9u8m5ndcsstwdZ0a7Nq+U870L3n5s2bozHVd9WqVcH290/1RX/P1DTC559/PthlSe3yEPEDAAAAAAAAAJAoPPgBAAAAAAAAAEgUUr0cGs41duzYYPv0A60S7ruoaFjtihUrgk0ayvb4EPFBgwYFe8iQIdGYpnrtvffewdZUELM4fUFDIn34paZH+DQjDZ1UW0M2zeLwzs997nPR2Mc//vFgazin70KgaQ+XXnppzWPsbeGcmv6j3WUWLlwYzdNOFUV3TUkdPcfqU2axv/jzqqHlGnLu52motYaV6/ppFndt0/Bps7hzhh6Tdkw0M7v++us7/dzehl8nVBPfnW3JkiXB1u5aPsz8vvvuC7be7/z7acj8sGHDorGDDz442JpqguIHNmwAABCRSURBVM8Wh+5ffLdRvbfqfcyn2mkYu0+j1mvL37uhNeh5HzlyZLAvuuiiaJ5qrOkHZuVNQagCtfaGzUDXRu2KaBZrqvN8dzctS+FTmlJP9VJ9fFre9773vWD7dU7vT+orfo+k6cu63/fofkQ7KprF90Ud0+8WZnHnKL/36W3fDbqLrpm+1Mbuu+8ebPUPX7pD/c9/51++fHmwq7C2EvEDAAAAAAAAAJAo7/jgJ8uyIVmWTc+ybFGWZQuzLPtqx88HZFn2YJZlyzv+u8c7vRe0DzRMgp3QsfqgYRLgiwmAhkmALyYAGiYBvpgAaJg2XYn4edPMJuV5PtrMxpjZl7IsG21mV5jZtDzPR5rZtI7XUF7QMA3QsfqgYRqgY/VBwzRAx+qDhmmAjtUHDRPmHWv85Hm+zszWddibsyxbbGb7mdknzezEjmn/ZWYzzOzyphxlC9Gcza9+9avB3mWXXaJ5kydPDrbWRjCLc0l9PRdF8w59PmG932uEPM/ndPy3VBr6mjnHH398sLUNpVlcj0RbVn7xi1+M5mmtHc231BpBZvH5923UtX3xs88+G2yf43vkkUcG2+fuautp1ddrq220ff6p5u6+9dZbfymrjkXgfUxrN+l1ovm0ZsW3pvW+qBThl2XScM899wx2vZxm/3cfc8wxwda8aO9jqofmuXt/U319jQJF/fKmm26KxubPn9/p5zaJJHxR687pfUxr+pjF513XVF/npdaaZxbXOdMaUTNmzIjmtTJHPgUNFb1/nHDCCdGY+pXWBNE6hGZxPbV6flSiuhJJ+GJX0ZbPkyZNCrav6aQ+W1ZcvZxeo2Ej+Fpo6rczZ84MttYsMYvPsa7BZvFeqqBaa6X1RX9f0fqfvv5qrXuc1g81M9u4cWOw662buvfxdZa0vmHfvn1rHpO+f7NptYa6bzCL94O69/Qa1rsH6f5D731au9fM7LOf/WywtV6l/yz1twceeCAa01pZRX93bwbdKu6cZdlQMzvKzJ4ws306HgqZma03s31q/M4lZnZJ44cIRYKGaYCO1QcN0wAdqw8apgE6Vh80TAN0rD5omCZdLu6cZVl/M7vLzC7L8zz637X5tsdunT56y/P8pjzPj83z/NgeHSn0GDRMA3SsPmiYBuhYfdAwDdCx+qBhGqBj9UHDdOlSxE+WZTvZtgvgZ3me393x4w1Zlg3O83xdlmWDzWxj7XcoLz7E7OKLLw62ph350K6777472L5NYK1QLx/6rq99io+GmRURbl0mDTV00qf3aHqXb7+u51VtDX02i1PCNHTSpyXUem+zuMXiEUccEWwf/qfvuWXLlmhMX+vv+XDR6667Ltj12ud2fF5pdCwCPX+aNmdmdtpppwVbW3r7EHZ/zspOuzXUc67r39y5c6N5el59Cpe2Gu3Xr1+w/TqmqC96v1e8j2mq5Xe+851g6xpstn36WLNpt46N4NcTPdfqV349rHWf8T/Xe9r73ve+aEzDqDVE3l9bPsS9mVRRw3rovU/vW2Zx+LzqPmvWrGheQSkfLSU1HRWfDj9u3Lhgn3jiicHWe6RZnF6ibd/N4hQfXefbmb6XsoZF4NdkTXG/8847g+1LDgwfPjzYPh1w9uzZwdZ9aU+ug7Lq6L9P6H6hqyU5NP3dLPajZcuWBVt9zzN06NDo9ebNm4O9bt26YKumZmbPPPNMl463CFqt4YgRI6LX+v3r+eefD/aqVauiebpX8HvPIUOGBPuss84K9tlnnx3N0/RHvS/6z7rtttuC/cQTT0Rjf/zjH4NdohTomnSlq1dmZrea2eI8z/9ThqaY2cQOe6KZ3et/F0oFGqYBOlYfNEwDdKw+aJgG6Fh90DAN0LH6oGHCdCXiZ5yZXWhmC7Ism9fxs2+Y2dVm9v+yLLvYzJ41s3Obc4hQEGhYffobOqYAGlYffDEN0LD64ItpgIbVB19MAzRMmK509XrUzGrlyJxc7OFAs8jzHA2rzxZ0rD5omAT4YgKgYRLgiwmAhkmALyYAGqZNt7p6pYi20jOLW4H3798/2I8//ng07+WXXw52V/Mtfe5fFXIBm4HWgfB1d5YsWRLsJ598MhobPHhwsDW/2b+H5jQPGjQo2L6WhOLrSuj7P/bYY8F+5JFHonmLFy8O9vr166MxzdetVzdBc+yr0AqwSFSTCy64IBpT7fT8aetEs+LPWep+qf6iucm/+MUvonkHHnhgsDXn2szs5JP/ev/XVpn1WrErXjPNv9ecbjOzf//3fw/2XXfdFWyfR5+6bs1Az1kRtV20HsmAAQOisVqt3uvVXoPuoXsWrZdnFmv96quvBlvvuX4etAf1j3r11XSe91+teXHcccdFY3PmzAm27lN8HRSuhfrUW7uafe5U79WrVwdb6/aYmQ0bNizYxxxzTDSm19LChQs7fe9UaHSfqDr6PYd+R9G6hd5ndS3etGlTNKa1QO+7775gP/XUU9G8Vta+awW6do0ZMyYa05b3a9euDbbXUN9D731mZqeeemqwzz///GBrTR//nlpP8tZbb43m/eY3vwm2r9NaNX/pclcvAAAAAAAAAACoFjz4AQAAAAAAAABIlF6Z6qXhmWeccUY0pukNGuKnrfTMGgsb9KGf9cLDekuIrQ+dfPjhh4OtrRLN4pQobQfs30PTTTSUddSoUdG8evpq+Lum9fnP0uugt2hWJBqqWc8/9Lz7VK+iSV1H/fv02tb0DzOzefPmBdu3X7/33r82dNCQ2qOOOiqapy2GNfxZfdnMbMWKFcGeNm1aNDZ16tRga2oalA/12blz50ZjqvnNN98cbG2rC91H9zO+bbeydevWYGvYuk89gPZQK23Ir5Wabq6toX2rbk0N8akJu+++e7BVf21nbJb+vbBoamnYjPOo76n3xQ0bNkTz+vbtG2z9fmNmdvjhhwdb99uppRUVhf/OpmlIU6ZMCfa73vWuaJ7ug3zbcd2DqS9WLX2ou+h16ff0Wl7jueeeC7ZPRVV/82uXlifQshH6uWbx+b/nnnuCrWl3ZvH+uOrrIhE/AAAAAAAAAACJwoMfAAAAAAAAAIBE6ZWpXhqGd9ZZZ0Vj2pVEq61rOJhZMaFeVQ8XaxRNj/KpG9oZS8Mo/e/VS7V78cUXg60pXDNnzuz2sUJzeeONN4L97W9/OxpTX9TwTA39NKMrUHfRdUdDZ30YraI6mcUhydoNxIfRdjX0XcN0fchu6iHPKaHXxZVXXllznqZ3oW/PUJ876KCDgr1y5cpo3qJFi4J9+eWXB9v7G7QHXRPrdRJSHa+66qpg+y5uus/1+yz1P//+UDz+Plj03l/3w2vWrInGdE3292dNgdF0bn+99NbvKp56+xZNp/R6a5dgT2/r5Ps2ut984IEHorGupvSrHn6PqvsKve79+V66dGmwf/KTnwTblz5IyQeI+AEAAAAAAAAASBQe/AAAAAAAAAAAJAoPfgAAAAAAAAAAEqVX1vjRXNY+ffpEY5onqC29fb58Svl+rUbPna/vQL2H3oVeC6tXr47GJk2aFGxth+nzpXtrjnQ7aaROEKSP1jzwLVqhGHbYIf7/dVq/YPHixcH+p3/6p2je7Nmzg7158+YmHR00G73fbd26tVPbrOu179jLNo4/d+06l7pvXrFiRTR2ww03BFtrgJmZzZo1K9jUeiqOslwXZUav2SK+9/ladfp9Qmv0an0zM7Orr7462FpXNmXNiPgBAAAAAAAAAEgUHvwAAAAAAAAAACRKr0z10pBGbd9mZvbYY48F+5577gm2tuoDgOLxKVvqcxqemXIIJgCAp17ajra+nTt3brB92iVt23sX3Cd7D7p38qksM2bMCPbMmTOjsVrpNlw7UDX894dp06YFe/DgwcFeuHBhNE99ordc90T8AAAAAAAAAAAkCg9+AAAAAAAAAAAShQc/AAAAAAAAAACJ0vIaP2/nqrc6l05z5LUd6u9+97toXv/+/YO9YcOGYPeW3L+u0C4NeyN6rRbdtvzt965CO3S91ny9iypch83yGXyxdTTTF6F1NMtnilxPfct2pU+fPjXHim6RW1b0HsDaB56y3xdrXb+N7m3q+YOuA77OV63aYd05jmbeF8uuY0o06/tAq75n+Puifn/Xer3Lly+P5qV0n+zqfZGIHwAAAAAAAACAROHBDwAAAAAAAABAomStDKHLsuxFM3vWzPYys5da9sGdU4ZjMGvNcRyY5/nAIt6oZBqa9a7jKFrHrdZ7zl1XqKKG+OL2VFFHfDGmihrii9tTRR3xxZgqaogvbk8VdcQXY6qoIb7YnmOoqWNLH/yED82y2XmeH9vyDy7ZMZTpOLpLWY6b42icshwzx9EzynLcHEfjlOWYOY6eUZbj5jgapyzHzHH0jLIcN8fROGU5Zo6jZ5TluMtwHGU4BlK9AAAAAAAAAAAShQc/AAAAAAAAAACJ0q4HPze16XOVMhyDWXmOo7uU5bg5jsYpyzFzHD2jLMfNcTROWY6Z4+gZZTlujqNxynLMHEfPKMtxcxyNU5Zj5jh6RlmOuwzH0fZjaEuNHwAAAAAAAAAAaD6kegEAAAAAAAAAJEpLH/xkWfbRLMuWZlm2IsuyK1r4uT/KsmxjlmVPy88GZFn2YJZlyzv+u0cLjmNIlmXTsyxblGXZwizLvtquY+kJvVlHNOzx56JhQbRLw47PRseCwBfRsIefjY4FgS+iYQ8/Gx0LAl9Ewx5+NjrWIs/zlvwzsz5mttLMhptZXzObb2ajW/TZE8zsaDN7Wn72LTO7osO+wsyuacFxDDazozvsXc1smZmNbsexoCMaoiEaomPv1RENq68hOqahIxpWX0N0TENHNKy+huj4DsfVQhHGmtnv5PWVZnZlCz9/qLsAlprZYBFnaStPfMfn3mtmp5ThWNARDdEQDdGxd+mIhtXXEB3T0BENq68hOqahIxpWX0N0rP2vlale+5nZ8/L6hY6ftYt98jxf12GvN7N9WvnhWZYNNbOjzOyJdh9LN0HHDtCwMNCw+5RNQzN0bISy6YiG3adsGpqhYyOUTUc07D5l09AMHRuhbDqiYfcpm4Zm6GhmFHc2M7N822O3vFWfl2VZfzO7y8wuy/P89XYeS0q08tyhYXNAwzRAx+qDhmmAjtUHDdMAHasPGqZBb9axlQ9+1pjZEHm9f8fP2sWGLMsGm5l1/HdjKz40y7KdbNsF8LM8z+9u57E0SK/XEQ0LBw27T9k0NEPHRiibjmjYfcqmoRk6NkLZdETD7lM2Dc3QsRHKpiMadp+yaWiGjmbW2gc/s8xsZJZlw7Is62tm55nZlBZ+vmeKmU3ssCfatty7ppJlWWZmt5rZ4jzP/7Odx9IDerWOaNgU0LD7lE1DM3RshLLpiIbdp2wamqFjI5RNRzTsPmXT0AwdG6FsOqJh9ymbhmbouI1WFhQys9NtW1XrlWb2zRZ+7u1mts7M/mLb8gwvNrM9zWyamS03s6lmNqAFxzHetoV0PWVm8zr+nd6OY0FHNERDNETH9v/DF9EQHcvxD19EQ3Qsxz98EQ3RsTn/so6DAwAAAAAAAACAxKC4MwAAAAAAAABAovDgBwAAAAAAAAAgUXjwAwAAAAAAAACQKDz4AQAAAAAAAABIFB78AAAAAAAAAAAkCg9+AAAAAAAAAAAShQc/AAAAAAAAAACJwoMfAAAAAAAAAIBE+f/+cPFid6qmDgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1440x288 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# view initial, noisy and decoded images\n",
    "plt.figure(figsize=(20, 4))\n",
    "plt.title('Reconstructed Images')\n",
    "\n",
    "print(\"Original Images\")\n",
    "for i in range(img_num_to_decode):\n",
    "    plt.subplot(2, img_num_to_decode, i+1)\n",
    "    plt.imshow(test_imgs_data_decode[i, ..., 0], cmap='gray')\n",
    "plt.show()    \n",
    "plt.figure(figsize=(20, 4))\n",
    "\n",
    "# plot the range of input noisy images\n",
    "print(\"Noisy Images\")\n",
    "for i in range(img_num_to_decode):\n",
    "    plt.subplot(2, img_num_to_decode, i+1)\n",
    "    plt.imshow(test_noisy_imgs_decode[i, ..., 0], cmap='gray')\n",
    "plt.show()    \n",
    "plt.figure(figsize=(20, 4))\n",
    "\n",
    "# plot the range of decoded noisy images\n",
    "print(\"Reconstruction of Noisy Images\")\n",
    "for i in range(img_num_to_decode):\n",
    "    plt.subplot(2, img_num_to_decode, i+1)\n",
    "    plt.imshow(decoded_images[i, ..., 0], cmap='gray')    \n",
    "plt.show()    \n",
    "plt.close()  "
   ]
  }
 ],
 "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}