mAiLab_0004_Edge_Detection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# HW0004 Edge Detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def read(datasets='train'):\n",
    "    prefix = \"datasets/\"\n",
    "    if datasets == 'train':\n",
    "        images_fname = prefix + \"train-images-idx3-ubyte\"\n",
    "        labels_fname = prefix + \"train-labels-idx1-ubyte\"\n",
    "    elif datasets == 'test':\n",
    "        images_fname = prefix + \"t10k-images-idx3-ubyte\"\n",
    "        labels_fname = prefix + \"t10k-labels-idx1-ubyte\"\n",
    "    images_file = open(images_fname, mode='rb')\n",
    "    labels_file = open(labels_fname, mode='rb')\n",
    "    \n",
    "    # Read image\n",
    "    images_file.seek(16)\n",
    "    dt = np.dtype((np.uint8, (28,28)))\n",
    "    images = np.fromfile(images_file, dtype=dt)\n",
    "    \n",
    "    #Read label\n",
    "    labels_file.seek(8)\n",
    "    dt = np.dtype(np.uint8)\n",
    "    labels = np.fromfile(labels_file, dtype=dt)\n",
    "    \n",
    "    return images, labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Basic"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Zero padding the first five images to 30 * 30, use Gx and Gy to filter them and print with reversed out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def padding(image, n):\n",
    "    return np.pad(image, [n, n], mode='edge')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def convolution(image, kernel):\n",
    "    image_x_size, image_y_size = image.shape\n",
    "    kernel_x_size, kernel_y_size = kernel.shape\n",
    "    conv_x_size, conv_y_size = image_x_size - kernel_x_size + 1, image_y_size - kernel_y_size + 1\n",
    "    conv = np.ndarray((conv_x_size, conv_y_size))\n",
    "    for i in range(0, conv_x_size):\n",
    "        for j in range(0, conv_y_size):\n",
    "            conv[i][j] = np.sum(np.multiply(image[i:i+kernel_x_size, j:j+kernel_y_size], kernel))\n",
    "    return conv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_images, train_labels = read()\n",
    "images = train_images[0:5]\n",
    "Gx = np.array([[-1, 0, 1], [-2, 0, 2], [-1, 0, 1]])\n",
    "Gy = np.array([[1, 2, 1], [0, 0, 0], [-1, -2, -1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXYAAABcCAYAAAB6Hp6BAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1053dad68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(len(images)):\n",
    "    plt.subplot(1, len(images), i + 1)\n",
    "    plt.imshow(convolution(images[i], Gx), cmap=plt.cm.gray)\n",
    "    plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXYAAABcCAYAAAB6Hp6BAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRhb2EELoGf8H36AOcvmXzIMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1053daba8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(len(images)):\n",
    "    plt.subplot(1, len(images), i + 1)\n",
    "    plt.imshow(convolution(images[i], Gy), cmap=plt.cm.gray)\n",
    "    plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Use Sobel filter kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def sobel_filter_kernel(size):\n",
    "    sobel = np.zeros((size, size))\n",
    "    for i in range(size):\n",
    "        for j in range(size):\n",
    "            sobel[i][j] = (size // 2) - j\n",
    "            if j > size // 2:\n",
    "                sobel[i][j] -= size // 2 - abs(i - size // 2)\n",
    "            elif j < size // 2:\n",
    "                sobel[i][j] += size // 2 - abs(i - size // 2)\n",
    "    return sobel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def print_image(image):\n",
    "    for i in image:\n",
    "        for j in i:\n",
    "            print(\"{:02X}\".format(j), end=' ')\n",
    "        print()\n",
    "    print()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 32 \\* 32 image with 5 * 5 sobel filter kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 FA D9 CA CD 0A A8 50 BF 1E 1B 8B E5 CC F5 6D FE 00 00 \n",
      "00 00 00 00 00 00 C4 9A 20 8C 69 C6 05 93 E9 C6 75 44 B5 A7 54 41 9E 92 A9 FD 00 00 \n",
      "00 00 00 00 00 9E 99 70 D8 0B 46 8C DF BC 1B E8 C0 6E 4B B9 85 00 4F D4 70 BC 00 00 \n",
      "00 00 00 00 00 49 94 6E A8 5E F4 7C 6F EC 96 B3 E2 C2 89 C6 DA D9 62 F2 24 7D 00 00 \n",
      "00 00 00 00 00 06 06 65 07 3B 0D C2 BC 2D 22 B8 82 B1 1E 38 7F 66 17 A6 D2 BE 00 00 \n",
      "00 00 00 00 00 25 06 B2 53 96 6E B7 FD B4 AB B0 70 12 08 EB CC 5E 12 9D 3B 80 00 00 \n",
      "00 00 00 00 00 68 FE E1 CE 29 11 31 01 E6 B2 BE 80 87 40 CC 86 CF DC 97 4E 00 00 00 \n",
      "00 00 00 00 00 DC 48 7F 74 B9 C7 39 C8 24 51 26 D0 B9 94 B0 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 60 4E 6F CE 0B C9 69 E4 3D 2A B7 D7 C9 36 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 E4 F0 FE EB 4F 32 16 D7 19 33 96 B7 0A 32 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 C9 9F F4 F0 F0 B8 CA B3 28 74 2B 92 36 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 EA 10 EB C3 C6 13 4F B4 0D 55 42 11 C7 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 BA 08 27 8C 17 C5 C3 EC 71 EC 54 CF 80 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 00 5E EC 1C 5B 9B B9 38 3B 6B 01 48 C4 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 B2 B1 BE 95 D7 B1 B0 A7 31 7A D2 E8 06 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 D0 04 5F 37 1C A5 40 2E D7 A3 E1 0A F5 45 C8 00 00 00 00 00 \n",
      "00 00 00 00 00 00 D2 65 CC C2 62 1C 11 7D 73 57 89 F5 F1 79 7C 7B 86 00 00 00 00 00 \n",
      "00 00 00 00 DC 98 5A 49 75 B7 55 A6 B4 09 62 01 28 02 63 F4 5B C2 04 00 00 00 00 00 \n",
      "00 00 92 71 5A 36 E0 6D 03 24 01 D6 9B E6 60 A1 8D 0C 14 F3 94 6C 00 00 00 00 00 00 \n",
      "00 00 4B 0C C3 62 C2 A9 42 28 BD 00 76 4C F9 CE 6C C1 BA E6 9C 00 00 00 00 00 00 00 \n",
      "00 00 8C A4 4D 05 FC FB 9A 99 78 DD 86 33 2F B7 C0 F8 A4 04 00 00 00 00 00 00 00 00 \n",
      "00 00 3B 02 F3 F8 7B 67 CD F9 EB E2 9D BB 59 F7 A9 12 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 FA 6A 9F 7A 3F F7 21 79 63 F9 8F 9D 15 16 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 F0 7E 09 91 68 15 68 2A 1F 92 18 20 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "\n"
     ]
    }
   ],
   "source": [
    "image = train_images[0]\n",
    "padding_image = padding(image, 2) # (32 - 28) / 2 = 2\n",
    "sobel = sobel_filter_kernel(5)\n",
    "conv_image = convolution(padding_image, sobel).astype(np.uint8)\n",
    "\n",
    "print_image(conv_image)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 34 \\* 34 image with 7 * 7 sobel filter kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 F7 C4 A3 94 53 72 A4 38 03 57 B0 A6 F2 1A E7 6A E3 7D 00 \n",
      "00 00 00 00 00 A6 58 80 52 82 2A E1 32 BE A0 71 48 F4 22 71 AF A0 0A 21 7B 22 BC 00 \n",
      "00 00 00 00 6D 5C 12 01 79 5D 94 89 21 CA ED 90 DD 8B 25 C6 97 06 77 FC DB 10 7B 00 \n",
      "00 00 00 00 06 6A F5 AD BB 0F A0 8E 53 AD AD 4B C4 D6 F4 73 26 60 EE FB A4 B0 3A 00 \n",
      "00 00 00 00 C3 50 39 87 8B A0 AC 92 F0 27 EF 59 04 C5 9E C4 E9 C8 E5 C2 83 64 FB 00 \n",
      "00 00 00 00 80 12 5F F4 1A 04 69 B1 17 02 27 53 D3 C4 65 D3 4E 22 26 25 78 12 3C 00 \n",
      "00 00 00 00 9F 12 D9 41 C9 9F CE BB EA 52 83 2A 92 DD FA 9A 1D C8 AF 26 AF 72 7D 00 \n",
      "00 00 00 00 E2 EC 27 38 37 89 6F 09 7B AB 3F E1 F3 0E 55 FA 40 75 5E DC F1 65 C0 00 \n",
      "00 00 00 00 25 A2 B3 BD 55 A0 AF 6E 4C 4D 6A CF 46 48 20 47 40 E6 D2 8D F6 75 00 00 \n",
      "00 00 00 00 CA 0B 9B D7 0E 81 E9 27 4C 60 20 4E 7E 8B E4 F2 D6 4B 00 00 00 00 00 00 \n",
      "00 00 00 00 00 10 54 09 17 B9 C7 AD 53 73 72 FD DA CE A6 E7 38 5C 51 00 00 00 00 00 \n",
      "00 00 00 00 00 00 D6 E1 BF 8B 73 75 E6 57 50 38 DB 35 F8 C8 52 93 9D 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 33 2B D6 4E 24 A9 F3 59 BE 0A 23 7D 21 3E DE C0 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 DF 24 C0 4D 6A 21 F0 7E 3E 56 CA 39 A1 5B 5E 06 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 00 22 99 0E 6A A8 7F F9 27 EB BE 55 6F A7 2A 48 00 00 00 00 \n",
      "00 00 00 00 00 00 00 B8 7A D1 25 A0 C0 21 50 88 FA 15 6D CC D6 C3 14 8A 00 00 00 00 \n",
      "00 00 00 00 00 BB 0C 86 0D 94 93 62 70 9E 08 60 72 FA 5D E0 81 A5 8C 4C 00 00 00 00 \n",
      "00 00 00 CA DB AB 5B 86 50 C7 62 04 99 09 21 9B DD A4 B8 F9 9D EF 2C 0A 00 00 00 00 \n",
      "00 5B 8E 83 B7 60 AB 6E 6A 99 C8 62 8B 8C 56 D0 0E 1E 0C FB A8 1C 3B C8 00 00 00 00 \n",
      "00 8C A4 33 81 21 8E 84 72 97 80 60 DD B6 70 11 41 C9 45 E8 3F 27 49 06 00 00 00 00 \n",
      "00 CD 3C DA 42 B5 7D 78 4F 1E A6 1D 27 42 74 18 87 92 13 73 B9 92 22 00 00 00 00 00 \n",
      "00 0E D4 A5 7D 79 72 F2 A0 B1 8F 24 B8 32 F9 94 68 3C F6 21 FF EA 00 00 00 00 00 00 \n",
      "00 BD 32 FA CE 87 8F 00 B3 2F 5F 56 0D DC 38 3B FA 2F 1A F7 06 00 00 00 00 00 00 00 \n",
      "00 7C 9A 53 0D 4F ED 6C 57 9D 5C 9B 10 3A 70 02 1E 02 1B 00 00 00 00 00 00 00 00 00 \n",
      "00 3B 02 D0 B4 33 2C B5 FD 84 2D 40 A2 A4 31 A5 21 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 68 F9 87 12 15 D3 21 2F 21 16 0E AD AC 30 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "\n"
     ]
    }
   ],
   "source": [
    "padding_image = padding(image, 3) # (34 - 28) / 2 = 3\n",
    "sobel = sobel_filter_kernel(7)\n",
    "conv_image = convolution(padding_image, sobel).astype(np.uint8)\n",
    "\n",
    "print_image(conv_image)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 36 \\* 36 image with 9 * 9 sobel filter kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 \n",
      "00 00 00 00 00 00 00 00 F4 AF 7C 5B 9C 33 B6 3C AC 4F C3 FC 9E 12 D6 4F 02 DF 59 FC \n",
      "00 00 00 00 88 16 E0 18 9B 34 DC 93 40 B2 29 A7 CD 7F 91 44 19 3C F9 53 AD E5 9B 7B \n",
      "00 00 00 3C 1F B4 48 0B 21 0C 43 FC 96 65 C4 46 18 48 41 DD 70 13 6A 3C F8 A4 B0 3A \n",
      "00 00 00 C3 40 AD DC 2C C5 BE 2B 96 EC 19 E6 74 A8 6A 53 AF 53 B1 CE 53 B6 6D 50 F9 \n",
      "00 00 00 80 D6 05 1C 7E FA 0C 0F CD 29 0A 39 D6 58 D3 CC 35 E9 E4 32 6A 74 36 F0 B8 \n",
      "00 00 00 3D 5C 99 F7 B2 BF 4E 15 89 33 32 3D 77 EA 9B 11 73 16 47 B8 15 FC 15 A4 79 \n",
      "00 00 00 FA 1E DB 31 0C A7 2F FD 14 5D 01 D8 29 CF 63 32 E5 1F A2 C6 04 A0 0A 52 BA \n",
      "00 00 00 19 1E 55 9D A5 09 35 7E 8D 29 CF B6 8E CA 3D 87 15 98 53 62 ED E2 41 B2 FB \n",
      "00 00 00 5C F8 A3 D7 03 F3 9A 92 85 34 2C D0 6D 01 C7 7F CF F4 26 FE D6 24 78 12 3C \n",
      "00 00 00 9F 72 C9 BF 09 51 D4 C4 12 96 8B 84 74 53 FF 61 E8 C7 58 17 26 7F 45 8F 00 \n",
      "00 00 00 E2 46 B3 45 0F 7F 79 73 D1 49 CD DF D9 FF BC 39 4E DD CC 6C 41 3E 55 9C 00 \n",
      "00 00 00 B8 CE B7 8A 31 55 01 5E 45 DF 27 F6 C0 A1 A4 E3 11 87 0B 15 73 00 00 00 00 \n",
      "00 00 00 00 C0 5A A3 84 FA 5F 35 50 6F E6 5F 7C 0E D4 F5 A3 A7 AD F6 ED 00 00 00 00 \n",
      "00 00 00 00 00 C8 D2 80 71 60 89 49 C9 5F 5A FE 14 91 0E F3 94 F0 13 3E 48 00 00 00 \n",
      "00 00 00 00 00 00 00 9D 1B A0 AE D5 E2 CC 8D 50 69 01 0A D2 5A 09 71 F6 8A 00 00 00 \n",
      "00 00 00 00 00 00 A0 C4 7B 8E FF 08 C7 C1 08 66 BA FC 7C FA CA 33 4B 56 CC 00 00 00 \n",
      "00 00 00 00 A4 B3 40 58 52 75 6F EB A7 38 18 8A 74 E2 C4 AB 8B 06 CB 40 0E 00 00 00 \n",
      "00 00 B8 1E FC 6D AE F5 2C 8C 44 4D 28 2B 77 D1 D3 1C DC 65 31 72 DF B8 D0 00 00 00 \n",
      "24 AB AC 38 F2 EF 20 A1 10 87 AD 20 BD 6F 25 93 17 A1 42 00 8C B5 0C 8E 8E 00 00 00 \n",
      "CD 3C DA 4C 7D 58 0C E7 26 23 32 6E C3 EE 65 8E 9C 78 8C 86 97 87 45 A7 4C 00 00 00 \n",
      "0E D4 81 4E 7B FB 0E F3 1F A9 D0 CF 4A A4 9C 1E B4 DA 97 73 A0 6E 1E FB 0A 00 00 00 \n",
      "4F 6C 28 50 A7 50 6C 55 3A 99 A4 0E F9 28 A4 12 EE 9D E1 81 31 D0 F3 D0 08 00 00 00 \n",
      "90 04 F3 CC 03 CF CA 37 C3 CD F2 61 93 79 13 83 8B 6E B6 3B 84 7F 90 D8 00 00 00 00 \n",
      "3F 62 48 CC 6F 92 4A 65 23 FD DC AF 85 87 63 67 5E 00 BB 29 88 18 38 00 00 00 00 00 \n",
      "FE CA A1 CA 71 EF C0 F3 B2 1D 6C 41 26 6C 07 48 FD 34 9E 3C 4A 08 00 00 00 00 00 00 \n",
      "BD 32 FA C8 B8 40 F0 54 0D A4 BC 02 65 49 6D 1A AB 45 5B 24 00 00 00 00 00 00 00 00 \n",
      "7C 9A 89 EB 9C 9C A4 95 6D F9 A6 25 BA 3B B9 C5 35 2C 00 00 00 00 00 00 00 00 00 00 \n",
      "\n"
     ]
    }
   ],
   "source": [
    "padding_image = padding(image, 4) # (36 - 28) / 2 = 4\n",
    "sobel = sobel_filter_kernel(9)\n",
    "conv_image = convolution(padding_image, sobel).astype(np.uint8)\n",
    "\n",
    "print_image(conv_image)"
   ]
  }
 ],
 "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}