reinforcing-learning.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 458,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import cv2\n",
    "import numpy as np\n",
    "from random import random\n",
    "import tensorflow as tf\n",
    "from random import random, seed\n",
    "from datetime import datetime\n",
    "\n",
    "CATEGORIES = [\"a\",\"b\",\"c\",\"d\",\"e\",\"f\",\"g\",\"h\",\"i\",\"j\",\"k\",\"l\",\"m\",\"n\",\"o\",\"p\",\"q\",\"r\",\"s\",\"t\",\"u\",\"v\",\"w\",\"x\",\"y\",\"z\"]\n",
    "\n",
    "path = \"C:/Users/user/Documents/Code/Machine-Learning/img\"\n",
    "pathSave = \"C:/Users/user/Documents/Code/Machine-Learning/imgToCharacters\"\n",
    "pathDataset = \"C:/Users/user/Documents/Code/Machine-Learning/my_dataset/train\"\n",
    "\n",
    "list_line = []\n",
    "list_y = []\n",
    "\n",
    "borderSize = 20\n",
    "\n",
    "IMSIZE = 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 459,
   "metadata": {},
   "outputs": [],
   "source": [
    "for img in os.listdir(path):\n",
    "    img_array = cv2.imread(os.path.join(path, img), 0)\n",
    "    ret, thresh = cv2.threshold(img_array, 127, 255, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 460,
   "metadata": {},
   "outputs": [],
   "source": [
    "kernel = np.ones((2,40),np.uint8)\n",
    "erosion = cv2.erode(thresh, kernel, iterations=10)\n",
    "border = cv2.copyMakeBorder(erosion, borderSize, borderSize, borderSize, borderSize, cv2.BORDER_CONSTANT, value=[255,255,255])\n",
    "\n",
    "image, contours, hierarchy = cv2.findContours(border, cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)\n",
    "\n",
    "for cnt in contours:\n",
    "        x,y,w,h = cv2.boundingRect(cnt)\n",
    "        if w*h < 600*3000 and w*h > 100*100:\n",
    "            if x < borderSize:\n",
    "                x = borderSize\n",
    "            if y < borderSize:\n",
    "                y = borderSize\n",
    "            img_crop = img_array[y-borderSize:y+h-borderSize, x-borderSize:x+w-borderSize]\n",
    "            list_line.append(img_crop)\n",
    "            list_y.append(y)\n",
    "            cv2.imwrite(os.path.join(pathSave, \"img-%s.png\" %random()), img_crop)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 461,
   "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"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.imshow(border, cmap=\"gray\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 462,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "17\n"
     ]
    }
   ],
   "source": [
    "array_y_line = []\n",
    "for i in range(len(list_y)):\n",
    "    array = [list_y[i], list_line[i]]\n",
    "    array_y_line.append(array)\n",
    "    \n",
    "array_y_line.sort(key=lambda x:x[0])\n",
    "\n",
    "print(len(array_y_line))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 463,
   "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"
    },
    {
     "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": "iVBORw0KGgoAAAANSUhEUgAAAXoAAABACAYAAAAUNrzvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAB4tJREFUeJzt3V+IHWcZx/Hvz2h6EQsmNkpIg6ayEHOlcQmBSm/ENM1N9EKIFxpqYW9SUNCL1d70UgW9EKQQMVBFDIKKvYjUEATxwpqNpPlj2GSN1a4JTUpFC4K1+ngx78Jhc/bsnN2ZM++85/eBw5l9d86Z9znvnGfmfWfOjCICMzMr1zu6roCZmbXLid7MrHBO9GZmhXOiNzMrnBO9mVnhnOjNzArXSqKXdETSoqQlSfNtLMPMzOpR0+fRS9oC3AA+CSwDF4DPRsQfG12QmZnV0sYe/UFgKSJuRcRbwBngWAvLMTOzGtpI9LuBVwf+Xk5lZmbWgXe28J4aUnbf+JCkOWAOYNu2bR/bt29fC1UxMyvXxYsXX4+InevN10aiXwb2DPz9MHB79UwRcQo4BTA7OxsLCwstVMXMrFyS/lJnvjaGbi4AM5L2StoKHAdeaGE5RZFU62FmNq7G9+gj4m1JTwMvAluA0xFxrenlmJlNg2E7eOOeLdnG0A0RcRY428Z7T7u19upbOE229ry+1LVZ3lpJ9DZ56yVmJ2Oz6eVEPyUkZZns6/Yccqy7Dec2zU/vEn2TByS7WtF8ULVd0zjs5ORqo2Sb6CeRDMddxrR8SXLcEOXaIymBNxLlyzbR56jOOHiOSbKujda9zuuaSBKrlzPpxDONPQUrgxN9g/qc5Ns2bZ+Ne4vrcy9tcpzobarlsgGaVG8hl3j7opRhLSf6KdLEDy9serWxB17KOpl77yTbO0zl/KFZ3iLC64/ZAO/RW+84iXdn1J6rh4XylXWi7/tZLNY8J/nybWY4x2dGDZd1os/BysrgDY5Zd1Z6Ek1+D1fea1IJfyN1b6puRSX6Nhts3Pfuw4ZhdUzuQVnOcl83c65f9om+r92raah3ziv2tPFG2kbJPtFbvnLv5Tj5Tdbg+lDK517KOrTu6ZWS9kj6taTrkq5J+mIq3yHpnKSb6Xl7Kpek70haknRZ0oG2g7B+WDntcdgjF6PquNFH08uflK6Xb82pcx7928CXI+LDwCHgpKT9wDxwPiJmgPPpb4AngJn0mAOea7zWVpw2kubK+9r9ctqg2HBNtsG6iT4i7kTEH9L0m8B1YDdwDHg+zfY88Kk0fQz4QVR+B7xH0q7GamxWQ1/vs7uZ+rbVy7D+G+uXsZI+CHwUeAl4f0TcgWpjALwvzbYbeHXgZcupzMymhDcmeamd6CW9G/gp8KWI+OeoWYeU3dfqkuYkLUhauHfvXt1qmFlPuEeRj1qJXtK7qJL8jyLiZ6n4tZUhmfR8N5UvA3sGXv4wcHv1e0bEqYiYjYjZnTt3brT+rRscAhj1MLP62jjo7Y3I2uqcdSPg+8D1iPj2wL9eAE6k6RPALwbKP5/OvjkE/GNliKcLTtTWNScf61qd8+gfBT4HXJF0KZV9Dfg68BNJTwF/BT6T/ncWOAosAf8Cnmy0xmPa7JfMX9J+yq3dcqtPyZr+rLtuuyaWv26ij4jfMnzcHeATQ+YP4OQm62VmZg1R11srAElvAotd12OCHgJe77oSEzZtMTvesuUS7wciYt2DnLlcAmExIma7rsSkSFqYpnhh+mJ2vGXrW7zZ3mHKzMya4URvZla4XBL9qa4rMGHTFi9MX8yOt2y9ijeLg7FmZtaeXPbozcysJZ0neklHJC2m69fPr/+KfpD0iqQrki5JWkhlxVzDX9JpSXclXR0oGzs+SSfS/DclnRi2rBysEe+zkv6W2viSpKMD//tqindR0uMD5b1Y35u8D0Uf2nhEvGW0cZvXnKhxTYotwJ+AR4CtwMvA/i7r1GBsrwAPrSr7JjCfpueBb6Tpo8AvqX6Ydgh4qev614jvMeAAcHWj8QE7gFvpeXua3t51bGPE+yzwlSHz7k/r8gPA3rSOb+nT+g7sAg6k6QeBGymuItt4RLxFtHHXe/QHgaWIuBURbwFnqK5nX6piruEfEb8B3lhVPG58jwPnIuKNiPg7cA440n7tx7dGvGs5BpyJiH9HxJ+pLgdykB6t79HcfSh60cYj4l1Lr9q460Rf8rXrA/iVpIuS5lJZ6dfwHze+EuJ+Og1VnF4ZxqCweLW5+1D0LuZV8UIBbdx1oq917fqeejQiDlDdWvGkpMdGzFvy5wBrx9f3uJ8DPgR8BLgDfCuVFxOvNn8fil7FPCTeItq460Rf69r1fRQRt9PzXeDnVF26TV3DvwfGja/XcUfEaxHx34j4H/A9qjaGQuJVM/eh6E3Mw+ItpY27TvQXgBlJeyVtBY5TXc++1yRtk/TgyjRwGLhKT67hvwnjxvcicFjS9tQlPpzKemHVcZRPU7UxVPEel/SApL3ADPB7erS+S43dh6IXbbxWvMW0cddHg6mO1t+gOlL9TNf1aSimR6iOtr8MXFuJC3gvcB64mZ53pHIB302fwRVgtusYasT4Y6qu7H+o9mKe2kh8wBeoDmQtAU92HdeY8f4wxXOZ6su8a2D+Z1K8i8ATA+W9WN+Bj1MNOVwGLqXH0VLbeES8RbSxfxlrZla4roduzMysZU70ZmaFc6I3MyucE72ZWeGc6M3MCudEb2ZWOCd6M7PCOdGbmRXu/24oXAk2lbjfAAAAAElFTkSuQmCC\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"
    },
    {
     "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"
    },
    {
     "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"
    },
    {
     "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"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAA8CAYAAABo1nFVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAB25JREFUeJzt3V2IHWcdx/Hvz2h7EQtNrEpIg6YSiLnSuKSBShHENM1N9EKIgoZa2JsEWlBwtTe9tIJeCFJYMWBFDIKKXkRqKIp40ZqNpHkxbLKN1W4TkpSKBgT74r8X8xw8bs7LnJfZmXnO7wOHc/bZOTPPf545/3nmmbPPKiIwM7N8vavuCpiZWbWc6M3MMudEb2aWOSd6M7PMOdGbmWXOid7MLHOVJHpJ+yUtS1qRtFDFNszMrBxN+3v0kjYAl4DPAKvAKeALEfGXqW7IzMxKqaJHvwdYiYgrEfEGcBw4WMF2zMyshHdXsM6twCtdP68C969dSNI8MA+wcePGT+zcubOCqpiZ5ev06dOvRcT7hy1XRaJXj7LbxociYhFYBJibm4ulpaUKqmJmli9JfyuzXBVDN6vAtq6f7wWuVrAdMzMroYoe/Slgh6TtwKvAIeCLFWxnZki9LpJ68yR17TNK+3arqq3L1sfHWntMPdFHxFuSjgLPAhuAYxFxYdrbsXr4pDNd4yb5qjStPjYdVfToiYgTwIkq1m2DSXKCbQEnVFtPlSR6M6tev5PFep3o124/1w5GDlexrU70o/aKqm6EHMc2m9rzHKdebdrvdWhqW9vkWpvo1/ugnOb2OuuqMvGM0wuZ9j514mgPt1XeWpvoxzErB/OocXpcf7JjY9b3nTXfTCX6NpqVk1NdprF/m9ZGddWnzHZzPynWfd+kHyf6BmtaAjGrWhX33fw5cqI3s5pNkoidxMtxojcg/w9M3ZfOZt0m/byNejw70ZuZVazujpT/laCZWeac6M3MMuehmyEiovbLrjoMGwOcxX1isyHHz7wT/QCdZFfXV7jGPeC661vVVAG+uWnWHq1N9E1LNFXVZ9L1Nm0/NU3Z/VN1D2/anYmqprWwdmptojdbT004YY5zhdeUExn0r0uOQyVNM/RmrKRtkn4n6aKkC5IeS+WbJZ2UdDk9b0rlkvQ9SSuSzkraXXUQZrMiIko9qlrvJI9p74NJ3191PZukzLdu3gK+GhEfBfYCRyTtAhaA5yJiB/Bc+hngYWBHeswDT0+91mY28+o+4bTJ0KGbiLgGXEuvb0m6CGwFDgKfSov9CPg98PVU/kwUe/R5SXdL2pLWY2bWeOt9Qqh66GqkMXpJHwY+DrwAfLCTvCPimqQPpMW2Aq90vW01lTnRm1lPs9rT7qg6/tJ/MCXpvcDPgccj4l+DFu1RdlsUkuYlLUlaunnzZtlqmJnZiEr16CW9hyLJ/yQifpGKr3eGZCRtAW6k8lVgW9fb7wWurl1nRCwCi2n9tyQtjxlDG90DvFZ3JdbZrMXsePPWlHg/VGahoYlexeDRD4GLEfHdrl/9GjgMfCs9/6qr/Kik48D9wD9LjM8vR8RcmQrnQNLSLMULsxez481b2+It06N/APgScE7SmVT2TYoE/zNJjwJ/Bz6ffncCOACsAP8GHplqjc3MbCRlvnXzR3qPuwN8usfyARyZsF5mZjYlTZm9crHuCqyzWYsXZi9mx5u3VsWrWf9ak5lZ7prSozczs4o40ZuZZa72RC9pv6TlNAnawvB3tIOklyWdk3RG0lIqy2YiOEnHJN2QdL6rbOT4JB1Oy1+WdLiOWMroE++Tkl5NbXxG0oGu330jxbss6aGu8lYc79OczLANbTwg3jzaeD1mrRswudAG4CXgPuAO4EVgV511mmJsLwP3rCn7NrCQXi8AT6XXB4DfUHy7aS/wQt31LxHfg8Bu4Py48QGbgSvpeVN6vanu2EaI90ngaz2W3ZWO5TuB7ekY39Cm4x3YAuxOr+8CLqW4smzjAfFm0cZ19+j3ACsRcSUi3gCOU0yKlquDFBPAkZ4/21X+TBSeB+5Of23cWBHxB+D1NcWjxvcQcDIiXo+IfwAngf3V1350feLt5yBwPCL+ExF/pfibkj206HiPiGsR8ef0+hbQPZlhdm08IN5+WtXGdSf6fhOg5SCA30o6LWk+lf3fRHDAsIng2mbU+HKI+2gaqjjWGcYgs3gHTWZIhm28Jl7IoI3rTvSlJkBrqQciYjfF/PxHJD04YNmc9wP0j6/tcT8NfAT4GMXsrN9J5dnEq8knM2xVzD3izaKN6070pSZAa6OIuJqebwC/pLiku94ZkhlnIrgWGDW+VscdEdcj4u2I+C/wA4o2hkzi1YDJDNPvs2rjXvHm0sZ1J/pTwA5J2yXdARyimBSt1SRtlHRX5zWwDzjP/yaCg9sngvty+ubCXspNBNdEo8b3LLBP0qZ0SbwvlbXCmvson6NoYyjiPSTpTknbKf7b2p9o0fEuDZ3MEDJq437xZtPGdd8Nprhbf4niTvUTdddnSjHdR3G3/UXgQicu4H0U/3bxcnrenMoFfD/tg3PAXN0xlIjxpxSXsm9S9GIeHSc+4CsUN7JWgEfqjmvEeH+c4jlL8WHe0rX8EyneZeDhrvJWHO/AJymGHM4CZ9LjQK5tPCDeLNrYUyCYmWWu7qEbMzOrmBO9mVnmnOjNzDLnRG9mljknejOzzDnRm5llzonezCxz7wC2d55tNk8OYgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAA8CAYAAABo1nFVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAB1RJREFUeJzt3V+IHWcZx/Hvr6ntRSw0aauENGhaAjFXNS4xUCmCmKa5iV4IsaChFvYmAQULrvamlyq0F4IUVgxYkQZBRS8iNRRFvGjNRtL8MWyyjbXdJiQpFQ0I1urjxbyLJ5uzZ+ecnTkz857fB5Yz5z1zzrzPeWeemXlnzruKCMzMLF+3NV0BMzOrlxO9mVnmnOjNzDLnRG9mljknejOzzDnRm5llrpZEL2mvpHlJC5Jm6liGmZmVo6rvo5e0DrgAfBZYBE4AX4yIP1e6IDMzK6WOI/pdwEJEXIqI94CjwP4almNmZiXUkeg3A2/1PF9MZWZm1oDba/hM9Sm7pX9I0jQwDbB+/fpPbN++vYaqmJnl6+TJk+9ExH2rzVdHol8EtvQ8vx+4vHymiJgFZgGmpqZibm6uhqqYmeVL0l/LzFdHoj8BbJO0FXgbOAA8XsNyaiP1Oym5lQeEq4e/f7NqVZ7oI+J9SYeBl4B1wJGIOFf1ctrCScnM2q6OI3oi4hhwrI7PbpOySd7qIck7ULMSakn0dRs2wToZ5KvfuuD2NrtZ5xL9KEfRS+9xArAuqbJbcJjtxttJfjqX6NdikrpavGFPjqq7sHyWlJ+JSvSTYtgdWlt2CpO0I+4aXw/pNif6MRnlukIbE98oG3wb48jNoHbx929O9JY9X7zPV1vORtvOid6yUsXGPGo3hZOOtZUTvXVeHV0WbejuGLb7rg11tnbqXKJva9+1NSPXI+Nc1/Hcz3ra+kv5ziX6SdHmDX153bq4QTat7RdPl9dN0tjrVfXyynzeWtblUc6+xrXtONGb2YrasNMZp1zjzf6fg/to08wm3UQc0fdL9mvtK8x1z28GXr9zMxGJvp+1Hum3/UdDvmhtZksmNtGPWxOJt+57wcsuwzsds2at2kcvaYuk30o6L+mcpK+m8o2Sjku6mB43pHJJ+p6kBUmnJe2sO4iuiIix/Y2rjk3HbtZV41x/y1yMfR/4ekR8DNgNHJK0A5gBXo6IbcDL6TnAY8C29DcNPF95rc1K8A7DrLBq101EXAGupOkbks4Dm4H9wKfTbD8Cfgd8I5W/EMXW84qkuyVtSp9TCW+YVrU2rVPj6HKzyTJUH72kjwIfB14FPryUvCPiiqQPpdk2A2/1vG0xlVWW6M3sZm3aUVn7lL6PXtIHgZ8BX4uIfwyatU/ZLWuhpGlJc5Lmrl+/XrYaZmY2pFJH9JI+QJHkfxIRP0/FV5e6ZCRtAq6l8kVgS8/b7wcuL//MiJgFZtPn35A0P2IMXXQv8E7TlRizSYvZ8eatLfF+pMxMqyZ6FZ1/PwTOR8RzPS/9CjgIfDs9/rKn/LCko8Angb+X6J+fj4ipMhXOgaS5SYoXJi9mx5u3rsVb5oj+YeBLwBlJp1LZtygS/E8lPQm8CXwhvXYM2AcsAP8Enqi0xmZmNpQyd938gf797gCf6TN/AIfWWC8zM6tIWwY1m226AmM2afHC5MXsePPWqXjl27LMzPLWliN6MzOrSeOJXtJeSfNpbJyZ1d/RDZLekHRG0ilJc6ksm/GBJB2RdE3S2Z6yoeOTdDDNf1HSwSZiKWOFeJ+R9HZq41OS9vW89s0U77ykR3vKO7G+VznGVRfaeEC8ebTxOAfa6jO+yDrgdeAB4A7gNWBHk3WqMLY3gHuXlX0XmEnTM8B30vQ+4NcUF713A682Xf8S8T0C7ATOjhofsBG4lB43pOkNTcc2RLzPAE/1mXdHWpfvBLamdXxdl9Z3YBOwM03fBVxIcWXZxgPizaKNmz6i3wUsRMSliHgPOEoxVk6u9lOMC0R6/FxP+QtReAW4O/0IrbUi4vfAu8uKh43vUeB4RLwbEX8DjgN766/98FaIdyX7gaMR8a+I+AvFrca76ND6HhFXIuJPafoG0DvGVXZtPCDelXSqjZtO9CuNi5ODAH4j6aSk6VR20/hAwGrjA3XNsPHlEPfh1FVxZKkbg8ziHTTGFRm28bJ4IYM2bjrRlxoXp6MejoidFMM2H5L0yIB5c/4eYOX4uh7388CDwEMUg/Y9m8qziVdrH+OqUzH3iTeLNm460ZcaF6eLIuJyerwG/ILilO7qUpfMKOMDdcCw8XU67oi4GhH/iYj/Aj+gaGPIJF4NGOMqvZ5VG/eLN5c2bjrRnwC2Sdoq6Q7gAMVYOZ0mab2ku5amgT3AWf4/PhDcOj7Ql9OdC7spNz5QGw0b30vAHkkb0inxnlTWCcuuo3yeoo2hiPeApDslbaX4Jzx/pEPru7TqGFeQURuvFG82bdz01WCKq/UXKK5UP910fSqK6QGKq+2vAeeW4gLuofhvXBfT48ZULuD76Ts4A0w1HUOJGF+kOJX9N8VRzJOjxAd8heJC1gLwRNNxDRnvj1M8pyk25k098z+d4p0HHusp78T6DnyKosvhNHAq/e3LtY0HxJtFG/uXsWZmmWu668bMzGrmRG9mljknejOzzDnRm5llzonezCxzTvRmZplzojczy5wTvZlZ5v4HBW3RLD3FjO8AAAAASUVORK5CYII=\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"
    },
    {
     "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"
    },
    {
     "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": [
    "array_line_word = []\n",
    "for y, line in array_y_line:\n",
    "    img_array = line\n",
    "    ret, thresh = cv2.threshold(img_array, 127, 255, 0)\n",
    "    kernel = np.ones((2,3),np.uint8)\n",
    "    erosion = cv2.erode(thresh, kernel, iterations=40)\n",
    "    border = cv2.copyMakeBorder(erosion, borderSize, borderSize, borderSize, borderSize, cv2.BORDER_CONSTANT, value=[255,255,255])\n",
    "\n",
    "    plt.imshow(border, cmap=\"gray\")\n",
    "    plt.show()\n",
    "    image, contours, hierarchy = cv2.findContours(border, cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)\n",
    "    list_word = []\n",
    "    list_x = []\n",
    "    for cnt in contours:\n",
    "        x,y,w,h = cv2.boundingRect(cnt)\n",
    "        if w*h < 1000*300 and w*h > 100*100 and h > 50:\n",
    "            if x < borderSize:\n",
    "                x = borderSize\n",
    "            if y < borderSize:\n",
    "                y = borderSize\n",
    "            img_crop = img_array[y-borderSize:y+h-borderSize, x-borderSize:x+w-borderSize]\n",
    "            list_word.append(img_crop)\n",
    "            list_x.append(x)\n",
    "    array_word_x = []\n",
    "    for i in range(len(list_x)):\n",
    "        array = [list_word[i], list_x[i]]\n",
    "        array_word_x.append(array)\n",
    "    \n",
    "    array_word_x.sort(key=lambda x:x[1])\n",
    "    array = []\n",
    "    for word, x in array_word_x:\n",
    "        array.append(word)\n",
    "    array_line_word.append(array)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 464,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "nombre de mot: 84\n"
     ]
    }
   ],
   "source": [
    "i = 0\n",
    "for line in array_line_word:\n",
    "    for word in line:\n",
    "        i += 1\n",
    "\n",
    "print(\"nombre de mot: \" + str(i))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 465,
   "metadata": {},
   "outputs": [],
   "source": [
    "array_line_word_char = []\n",
    "\n",
    "for line in array_line_word:\n",
    "    \n",
    "    array_word_char = []\n",
    "    \n",
    "    for word in line:\n",
    "        \n",
    "        word_char = []\n",
    "        \n",
    "        img_array = word\n",
    "        ret, thresh = cv2.threshold(img_array, 127, 255, 0)\n",
    "        kernel = np.ones((45,3),np.uint8)\n",
    "        erosion = cv2.erode(thresh, kernel, iterations=1)\n",
    "        border = cv2.copyMakeBorder(erosion, borderSize, borderSize, borderSize, borderSize, cv2.BORDER_CONSTANT, value=[255,255,255])\n",
    "        #plt.imshow(border,cmap=\"gray\")\n",
    "        #plt.show()\n",
    "        image, contours, hierarchy = cv2.findContours(border, cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)\n",
    "        list_char = []\n",
    "        list_x = []\n",
    "        \n",
    "        for cnt in contours:\n",
    "            x,y,w,h = cv2.boundingRect(cnt)\n",
    "            if w*h < 150*100 and w*h > 20*50:\n",
    "                if x <= borderSize:\n",
    "                    x = borderSize\n",
    "                if y <= borderSize:\n",
    "                    y = borderSize\n",
    "                img_crop = thresh[y-borderSize:y+h-borderSize, x-borderSize:x+w-borderSize]\n",
    "                list_char.append(img_crop)\n",
    "                list_x.append(x)\n",
    "        \n",
    "        # Classer\n",
    "        \n",
    "        array_char_x = []\n",
    "        for i in range(len(list_char)):\n",
    "            array = [list_char[i], list_x[i]]\n",
    "            array_char_x.append(array)\n",
    "            \n",
    "        array_char_x.sort(key=lambda x:x[1])\n",
    "        \n",
    "        array = []\n",
    "        for char, x in array_char_x:\n",
    "            array.append(char)\n",
    "        array_word_char.append(array)\n",
    "        \n",
    "    array_line_word_char.append(array_word_char)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 492,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = tf.keras.models.load_model(\"readable-model.model\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 493,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "la volonte du vrai qui noug entrai encore dans nombre d entreprises perilleuses cette celebte veracite dent jusgu ici tous les philosophes ont parle avec veneration que de problemes pleins d equivogues c est deja une longue histoire et pourtant semble c il elle vient justement de commencer guoi d etonnant que finalement noos devions mepiants perdions patience et nous detournions excedes cz sphinx ne nous dpprendra t il pas a nous aoss i de notre cote l art d interroger gui est cz\n"
     ]
    }
   ],
   "source": [
    "array_predictions = []\n",
    "\n",
    "for line in array_line_word_char:\n",
    "    for word in line:\n",
    "        for char in word:\n",
    "            new_char = cv2.resize(char, (IMSIZE, IMSIZE))\n",
    "            new_char = np.array(new_char).reshape(-1,50,50,1)\n",
    "            new_char = new_char / 255.0\n",
    "            prediction = model.predict(new_char)\n",
    "            array_predictions.append(CATEGORIES[np.argmax(prediction)])\n",
    "        array_predictions.append(' ')\n",
    "\n",
    "text = \"\".join(array_predictions)\n",
    "text = text.strip()\n",
    "print(text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 494,
   "metadata": {},
   "outputs": [],
   "source": [
    "real_text = \"la volonte du vrai qui nous entrai encore dans nombre d entreprises perilleuses cette celebre veracite dont jusqu ici tous les philosophes ont parle avec veneration que de problemes pleins d equivoques c est deja une longue histoire et pourtant semble t il elle vient justement de commencer quoi d etonnant que finalement nous devions mefiants perdions patience et nous detournions excedes ce sphinx ne nous apprendra t il pas a nous aussi de notre cote l art d interroger qui est ce\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 495,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "484 483\n"
     ]
    }
   ],
   "source": [
    "print(len(text.strip()), len(real_text))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 496,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "21 g s\n",
      "77 t r\n",
      "88 e o\n",
      "94 g q\n",
      "166 g q\n",
      "211 c t\n",
      "243 g q\n",
      "271 o u\n",
      "282 p f\n",
      "329 z e\n",
      "342 d a\n",
      "363 o u\n",
      "393 g q\n",
      "400 z e\n",
      "14\n"
     ]
    }
   ],
   "source": [
    "miss = []\n",
    "i = 0\n",
    "text = text.replace(\" \", \"\")\n",
    "real_text = real_text.replace(\" \", \"\")\n",
    "\n",
    "#if len(text) == len(real_text):\n",
    " #   i = 0\n",
    "for line in array_line_word_char:\n",
    "    for word in line:\n",
    "        for char in word:\n",
    "            if text[i] != real_text[i]:\n",
    "                print(i, text[i], real_text[i])\n",
    "                miss.append([char, real_text[i]])\n",
    "            i += 1\n",
    "                \n",
    "#else:\n",
    " #   print(\"Erreur mauvaise correction\")\n",
    "\n",
    "print(len(miss))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 490,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "c\n",
      "t\n",
      "t\n",
      "u\n",
      "u\n",
      "v\n",
      "t\n",
      "c\n",
      "s\n",
      "s\n"
     ]
    }
   ],
   "source": [
    "for img, cat in miss:\n",
    "    print(cat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 491,
   "metadata": {},
   "outputs": [],
   "source": [
    "seed(datetime.now())\n",
    "\n",
    "for img, cat in miss:\n",
    "    t = random()\n",
    "    pathCat = os.path.join(pathDataset, cat)\n",
    "    cv2.imwrite(os.path.join(pathCat,\"img%s.png\"%t, ), img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.6 (tensorflow)",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}