{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras.datasets import mnist\n",
    "import numpy as np\n",
    "from keras.utils.np_utils import to_categorical\n",
    "from keras.layers import Flatten,LeakyReLU\n",
    "from keras.layers import Activation,BatchNormalization\n",
    "from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D,AveragePooling2D,Reshape, Conv2DTranspose\n",
    "from keras.models import Model\n",
    "from keras import backend as K\n",
    "from scipy import misc\n",
    "from os import listdir\n",
    "import scipy\n",
    "import scipy.ndimage as ndimage\n",
    "# import cv2\n",
    "from scipy.misc import toimage\n",
    "import scipy.ndimage\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def LAPSRN(x):\n",
    "    temp = Conv2D(64,(3,3),padding='same')(x)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU(alpha=0.2)(temp)\n",
    "    temp = Conv2D(64,(3,3),padding = 'same')(temp)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU(alpha=0.2)(temp)\n",
    "    temp = Conv2DTranspose(1,(4,4))(temp)\n",
    "    x2 = temp\n",
    "    temp = Conv2D(64,(3,3),padding = 'same')(temp)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU(alpha = 0.2)(temp)\n",
    "    temp = Conv2D(64,(3,3),padding = 'same')(temp)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU(alpha=0.2)(temp)\n",
    "    temp = Conv2DTranspose(1,(4,4))(temp)\n",
    "    x4 = temp\n",
    "    temp = Conv2D(64,(3,3),padding = 'same')(temp)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU()(temp)\n",
    "    temp = Conv2D(64,(3,3),padding = 'same')(temp)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU()(temp)\n",
    "    temp = Conv2DTranspose(1,(4,4))(temp)\n",
    "    x8 = temp\n",
    "    temp = Conv2D(64,(3,3),padding = 'same')(temp)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU(alpha = 0.2)(temp)\n",
    "    temp = Conv2D(64,(3,3),padding = 'same')(temp)\n",
    "    temp = BatchNormalization(temp)\n",
    "    temp = LeakyReLU(alpha = 0.2)(temp)\n",
    "    temp = Conv2DTranspose(1,(3,3))(temp)\n",
    "    x16 = temp\n",
    "    return (x2,x4,x8,x16)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def preprocess():\n",
    "    (x_train,y_train),(x_test,y_test) = mnist.load_data()\n",
    "    x_train = x_train.astype('float32') / 255.\n",
    "    x_test = x_test.astype('float32') / 255.\n",
    "\n",
    "    print (x_train[0][0])\n",
    "\n",
    "    _y_train = np.zeros(shape=(x_train.shape[0], 39, 39))\n",
    "    _y_test = np.zeros(shape=(x_test.shape[0], 39, 39))\n",
    "    \n",
    "    for i in range(len(y_train)):\n",
    "        _y_train[i, :] = scipy.ndimage.zoom(x_train[i, :], 1.4, order=1)\n",
    "\n",
    "    for j in range(len(y_test)):\n",
    "        _y_test[j, :] = scipy.ndimage.zoom(x_test[j, :], 1.4, order=1)\n",
    "\n",
    "    x_train = ndimage.gaussian_filter(x_train,0.9)\n",
    "    x_test = ndimage.gaussian_filter(x_test,0.9)\n",
    "    \n",
    "    x_train = np.reshape(x_train, (len(x_train), 28, 28, 1))\n",
    "    x_test = np.reshape(x_test, (len(x_test), 28, 28, 1))\n",
    "    \n",
    "    _y_train = np.reshape(_y_train, (len(_y_train), 39, 39, 1))\n",
    "    _y_test= np.reshape(_y_test, (len(_y_test), 39, 39, 1))\n",
    "    return (x_train,_y_train,x_test,_y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "input_img = Input(shape = (28,28,1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "op1,op2,op3,op4 = LAPSRN(input_img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model1 = Model(input_img,op1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model2 = Model(input_img,op2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model3 = Model(input_img,op3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model4 = Model(input_img,op4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model4.compile(optimizer='adadelta',loss = 'categorical_crossentropy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.\n",
      "  0.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n"
     ]
    }
   ],
   "source": [
    "x_train,y_train,x_test,y_test = preprocess()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28, 1)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.shape(x_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 39, 39, 1)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.shape(y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "138s - loss: 0.0579\n",
      "Epoch 2/10\n",
      "71s - loss: 0.0431\n",
      "Epoch 3/10\n",
      "71s - loss: 0.0361\n",
      "Epoch 4/10\n",
      "72s - loss: 0.0327\n",
      "Epoch 5/10\n",
      "73s - loss: 0.0310\n",
      "Epoch 6/10\n",
      "73s - loss: 0.0300\n",
      "Epoch 7/10\n",
      "73s - loss: 0.0292\n",
      "Epoch 8/10\n",
      "71s - loss: 0.0285\n",
      "Epoch 9/10\n",
      "71s - loss: 0.0279\n",
      "Epoch 10/10\n",
      "71s - loss: 0.0274\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7fc92a878e80>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model4.fit(x_train,y_train,batch_size=256,shuffle=True, verbose=2, epochs=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "out = model4.predict(np.expand_dims(x_test[0], axis=0))\n",
    "out2 = np.reshape(out, (39, 39))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhEQKg5+QSGHwExIpDH5CIuX/AOkVTBkDtKIeAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc92ab4f9b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#reduced noise image\n",
    "plt.imshow(toimage(out2))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nP5FT/w+Snwk/x1a+twAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc8e818c278>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#input image\n",
    "img2 = np.reshape(x_test[0], (28, 28))\n",
    "plt.imshow(toimage(img2))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "img3 = scipy.ndimage.zoom(x_test[0], 1.4, order=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc92ab27400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "img3 = np.reshape(img3, (39, 39))\n",
    "plt.imshow(toimage(img3))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fc8e0657470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x_test20 = ndimage.gaussian_filter(x_test2[0],0.9)\n",
    "img3 = scipy.ndimage.zoom(x_test20, 1.4, order=3)\n",
    "img3 = np.reshape(img3, (39, 39))\n",
    "plt.imshow(toimage(img3))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "(x_train2,y_train2),(x_test2,y_test2) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_train5 = ndimage.gaussian_filter(x_train2,0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_train9 = ndimage.gaussian_filter(x_train2,0.9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "img1 = np.reshape(x_train5[0], (28, 28))\n",
    "img2 = np.reshape(x_train9[0], (28, 28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f35f8077cf8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(toimage(img1))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f35f06aa9e8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(toimage(img2))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12+"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}