CNN_RPS_keras.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import keras\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, Convolution2D, MaxPooling2D, Flatten, Dense"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_datagen = ImageDataGenerator(\n",
    "        rescale=1./255,\n",
    "        shear_range=0.2,\n",
    "        zoom_range=0.2,\n",
    "        horizontal_flip=False)\n",
    "\n",
    "test_datagen = ImageDataGenerator(rescale=1./255,\n",
    "        shear_range=0.2,\n",
    "        zoom_range=0.2,\n",
    "        horizontal_flip=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 2188 images belonging to 3 classes.\n",
      "Found 2188 images belonging to 3 classes.\n"
     ]
    }
   ],
   "source": [
    "train_generator = train_datagen.flow_from_directory(\n",
    "        'data',\n",
    "        target_size=(28, 28),\n",
    "        batch_size=32,\n",
    "        class_mode='binary')\n",
    "\n",
    "test_generator = test_datagen.flow_from_directory(\n",
    "        'data',\n",
    "        target_size=(28, 28),\n",
    "        batch_size=32,\n",
    "        class_mode='binary')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'paper': 0, 'rock': 1, 'scissors': 2}"
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_generator.class_indices"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1444d4160>"
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAFpNJREFUeJztnV2MnOV1x/9nZvZ77cU2YJuPYiCohaDGqVYoUlCVKk1EUCTIDQoXkSOhOBdBaqRcFNGLcomqJhEXVSSnWDFVSlIpQXCB2iSoEkpVRSyE8hHaQsAUG+P113pnv2Z25j292IEusM//zO7Mzoz7/H+S5dk588z7zDvv/31n5v+cc8zdIYTIj1K/JyCE6A8SvxCZIvELkSkSvxCZIvELkSkSvxCZIvELkSkSvxCZIvELkSmVXm6sNF728lR6k2bGn4CEo6FsbJsPuCQx8BWc276+k20gWF0azq2TyXe8spUfL+HKWRKOx6bjRbVAsVK0dTB3JH4zuwPAIwDKAP7e3R9mjy9PVbDn61cl46XhMt1ehcQteCVW5vvDjH8IoiemDj8/RQLt5Ci3go8tOhWB8/1aNJvpoc0iGMvjHrw2kHjRjMbysIWvO9jv9UYy1iQxAGjUV5OxhSerdOx6tnzYmlkZwN8B+BKAWwDca2a3bPX5hBC9pZNr1m0A3nD3N929DuAnAO7qzrSEENtNJ+K/GsA76/4+0brvQ5jZYTObMbOZYin9EVAI0Vu2/dd+dz/i7tPuPl0a59/phRC9oxPxnwRw7bq/r2ndJ4S4BOhE/M8BuMnMrjezYQBfBfBUd6YlhNhutmz1uXvDzO4H8C9Ys/qOuvurdJABVkmfb8pD/FxUJt8arBxYdZXAl6VRgDmB1ukagXCRwtY942hqFnrtwbYjy4yMbxIbEACKRvAbUWD1MSuwCMZGVp55dN0M9gt7zwPbuVtLUjry+d39aQBPd2cqQoheouW9QmSKxC9Epkj8QmSKxC9Epkj8QmSKxC9EpvQ0n9/MUB5Km5Tl4ciUTsdLkY9f4r5rKfRW0/GwDkFk0xvPH+1kHUGUG26RTR+khjc98OrZOoBgjUCwW8K0WyfPH6U6R+nCRcFft4dzS8eazXTK7tq200++mQ5cuvILkSkSvxCZIvELkSkSvxCZIvELkSkSvxCZ0lOrD+CWXFRht1Qi56qgSJCxsW3EO8mjDKuKRzZjJ9V9oyrOURHb1cDKW+G2lDfStpRFVl8Qj6r/Mjsveu4o5TesHNzooDrwalC1mMU3UYxZV34hMkXiFyJTJH4hMkXiFyJTJH4hMkXiFyJTJH4hMqXHKb3c5y8FlnSJeflhRm7k8wcpwSQclbcOy4IHKwHiwt5k7UQwtiA+PAA06nUab67wOEurjVJ6PSjdHZX2ptuOKmsHpbvDsuHhOoJ0LOrS66RL72Zaj+vKL0SmSPxCZIrEL0SmSPxCZIrEL0SmSPxCZIrEL0SmdOTzm9lxAFUATQANd5/mA3iL7shrZ224LSjNHZ7mwqR7EgpLd28iyXqzG1/bQHpkMLTZ4J5yfblG48Uy9/kP2FQytrPBD79x4/Fmwee+spr2w+uBl14L2ofPOX/ds40FGm8Qo78RrW+okdLdYcv0/6Mbi3z+zN3PduF5hBA9RB/7hciUTsXvAH5hZs+b2eFuTEgI0Rs6/dh/u7ufNLMrAfzSzP7T3Z9d/4DWSeEwAJR39bxkoBAiQUdXfnc/2fp/FsATAG7b4DFH3H3a3afLE0GVTSFEz9iy+M1swsx2vH8bwBcBvNKtiQkhtpdOPofvBfBEy+aqAPhHd//nrsxKCLHtbFn87v4mgE9tapAZSqw2fyc+f9BrOqx8H/WqJoZ56PNH8ZCoxzcJBb2io3z8UpXHbx66ksY/VU7Hb7ycjx0Leg6sBrUG5qqLydi5uTk+tr5M46cbSzQ+EdTeL0h9iaiUwHJjJRl7B+nX/FFk9QmRKRK/EJki8QuRKRK/EJki8QuRKRK/EJnS29LdiNpsB1YfK58dZdWGVl5U+5uH+djICuzguQEYcZUaUYvtBW6X7V0eofHbJ/+Axm+YvDwZO3DVNXRsfYnbbZHVt3O0moyNlfmhP7XArbypxVEa39Pgq1nHxieTsdFR/tzVynwy9mT5DTp2PbryC5EpEr8QmSLxC5EpEr8QmSLxC5EpEr8QmSLxC5EpPa+rRVNvA6/dmV8elt7urLQ3T9uN1idE59i4iTcdTdJ2PShRXVrm6wD+aHQ/jQ8FKb9Te8aSseYKLwverPF4VBJ9YjS97UqQTrxznK8xGDrLC1b7Kt8vFVKWfPdoeg0AAOwbTr+ukcrbdOx6dOUXIlMkfiEyReIXIlMkfiEyReIXIlMkfiEyReIXIlN67vM78azjtPa0rxvl88fltaPx23me3HoLbgBAk/j8K9znH+Z2NG6c3EXjpYV0G2yAX10qrLYDgOUmn3vEUGUoHRsapmOLoNX1WDB+z+ROGq+SWgUl0r4bABDF20RXfiEyReIXIlMkfiEyReIXIlMkfiEyReIXIlMkfiEyJfT5zewogC8DmHX3W1v37QbwUwAHABwHcI+7Xwi3ZoDR2vxBm22avx3UAqBRwDo4D4ZrCAKCtHR4wR/QqKe99sYSz4mfbPJDYKyI6s9zv3vnzqlk7LrrP0HHnj97hsbnqxdpfLWW3i8rKzxfH87rHExMjNN4KagfsWMinbM/NMTfk7kL59PBoCX7eto54n8E4I6P3PcAgGfc/SYAz7T+FkJcQoTid/dnAXz0VHMXgGOt28cA3N3leQkhtpmtftbd6+6nWrffA7C3S/MRQvSIjn/w87Uv4skvOGZ22MxmzGymucC/RwkhesdWxX/azPYDQOv/2dQD3f2Iu0+7+3R5kv94JIToHVsV/1MADrVuHwLwZHemI4ToFaH4zexxAP8O4A/N7ISZ3QfgYQBfMLPXAfx5628hxCVE6PO7+72J0Oe3tsnIcWcwP537myV09pXDO5h3qcO6/UXgOddZ/ftFnhN/YPwqGt8/xfP5p0ifeQDYuSOd1x7l1FeGeby+skLjq430ay+V+NqMyGsfHR0JxvPjrVJJP/+Fc7wnwGVT6X1aLrd/nGuFnxCZIvELkSkSvxCZIvELkSkSvxCZIvELkSm9Ld3tvCRyiab78tTZKC3WAmsHYWlvUnI8bA8ehYMW3EEZadTSVuBogz/3ZYHdFpWonhzfQeNs5tVqlY6t1Xhd8ZHxCRofLaevbeVyuqw3ACwuLNB48/w5Go/e0wXy2peWlujYHTv4Pm8XXfmFyBSJX4hMkfiFyBSJX4hMkfiFyBSJX4hMkfiFyJSet+imdFABO6iUHJT9BsLzIB3fYenuoOWyN3hKb5msA5gq8dTTq8Yuo/GdUzy+uDBP4w1SInthko+tVHh6qq/ysuSNWnq/FKR9NwCsLAY+/ypPlZ6aSpcsB4C5C+m03YlxXhZ8ZDg9982UkdeVX4hMkfiFyBSJX4hMkfiFyBSJX4hMkfiFyBSJX4hM6a3Pb4CV0ucbL7ifTZ32wN8smtxLD91RUmsgWmMQnWGjua2u8Lz28Vp6bgeHeBvFz+y/icZHjB8ii3U+t52j6dLeV++/ho51cC99aWmRxldX0y26o3clek9qwXuyUOXrBFhJ81o9WL9A5raZAvO68guRKRK/EJki8QuRKRK/EJki8QuRKRK/EJki8QuRKaHPb2ZHAXwZwKy739q67yEA3wBwpvWwB9396XBrznPXS1F7YWLGRz5/5H9GLbiN2L7R2CJo0V00+Hhf5Z5zxdPPf8U4zysfjeYWeM779u2j8cnRdI35y6/YT8eePXuKxms15uMDdTL36jzvGdAI8vV3X8b369Iyr71/Zv5CMlYKWrJXSPvwzVzN23nsjwDcscH933f3g61/sfCFEANFKH53fxbA+R7MRQjRQzr5zn+/mb1kZkfNbFfXZiSE6AlbFf8PANwI4CCAUwC+m3qgmR02sxkzm2ku8O9RQojesSXxu/tpd2+6ewHghwBuI4894u7T7j5dnhyseqFC5MyWxG9m63+m/QqAV7ozHSFEr2jH6nscwOcAXG5mJwD8NYDPmdlBrDloxwF8cxvnKITYBkLxu/u9G9z96DbMJYbY4ZHXHiXsO7fS6WekuFZ64OODbzya2nItXRv/3XPv8cHX3EzDo+OjNH7xLDeCVsfSXvuOqcjHX6HxoRE+t+ER0rMgOFzmLqR9eACoVi/SeL3O1yAsL6bXGYxPTNCxSwvpOgZF0ANiPVrhJ0SmSPxCZIrEL0SmSPxCZIrEL0SmSPxCZErvl9wVpG1yYGpRS63UWZvssEM38wqjU2jQHtyDMtFRC+9dE+ky0Af3fIKOnb/ILasTs+do/OZPfpLGC7Kie2mJW3nlYb5jh0u8zTaz62orfNsFLfsNnDp5ksZXopLmU+lU52bwfs8tpG3CRlD+fj268guRKRK/EJki8QuRKRK/EJki8QuRKRK/EJki8QuRKT31+d0dTeJDliwo3U2fPCrdHeRwBuONrE/wDlN6m6vcm20scc+4vDqcjF2xK70GAAB2T1xG41fu4aW5xyaC0uDj6fTUiZ18bL3Gy18vEr8b4OtCePtuYGmRt/+u13lJuosk7RYAzq+k4789/hYd++piOoX7bLB2Yj268guRKRK/EJki8QuRKRK/EJki8QuRKRK/EJki8QuRKb3P52eWtgV570GcYaUoYT/IqSf5/FZ0dg6NWnA3A59/hbSDro4u0LH7JnbT+PgYL4/dbHC/u9lIz318dIyOXSVeOAAszPGy4XPn0/n81Xlex6BW463JfYivSXlraZ7Gf7uYLlt+tpz28QHgIlmasVpuXyO68guRKRK/EJki8QuRKRK/EJki8QuRKRK/EJki8QuRKaHPb2bXAngMwF6sJaYfcfdHzGw3gJ8COADgOIB73J33NQbg1E+P8vlJ3nxQGz9sXRyl5HewxiAa2WwG+fxBvn91Pp3X/tYSzw2/apzn8w8FNRZGRkkbbABAOp+/GeyY2hL3+Rfn5mi8uZpeY1AZ4jX/a0Hd/eoyz5t/d4GvIzgzll4f0ajwfTpaSR+s9QpfI7Cedq78DQDfcfdbAHwGwLfM7BYADwB4xt1vAvBM628hxCVCKH53P+XuL7RuVwG8BuBqAHcBONZ62DEAd2/XJIUQ3WdT3/nN7ACATwP4DYC97v7+GsX3sPa1QAhxidC2+M1sEsDPAHzb3T+0cNndHYmvtmZ22MxmzGymudB+HzEhxPbSlvjNbAhrwv+xu/+8dfdpM9vfiu8HMLvRWHc/4u7T7j5dnuygQKcQoquE4re1EqiPAnjN3b+3LvQUgEOt24cAPNn96Qkhtot2Uno/C+BrAF42sxdb9z0I4GEA/2Rm9wF4G8A98VMlvx2sRUM3bet2W0FKbwOARVYebQEepANHnlZgQ45V0qW5AWDveDrtdv/oHjr24jnuzhZBieqJCZ6WWy6nLbXq+Q0/LH7AubM8Xmvw8tvLxOqrB/bq20vcRvy389xCfWecpwQ3xtLS83LwCblEjpdNdKoPxe/uvyZP+fn2NyWEGCS0wk+ITJH4hcgUiV+ITJH4hcgUiV+ITJH4hciU3pbudl4h25tB2i3z2jfhb2647WgJAZm4F3zjkVfudf66r9txJY3fse/mZOyPd9xAx1qQLlxv8tTW02fepfG3Tp1IxkZHx+nY5eWgTXbg8w+Ppdcg1IM1I28u8/UPv6/wuS0PBy3jK2nplSqBz0/WpBhdj/KR7bT9SCHE/yskfiEyReIXIlMkfiEyReIXIlMkfiEyReIXIlN66vM7Ap/fAp+flu4Ohgb+Z1wpgD0iKBveCF5XsPE9Yztp/Prx9DqAapX70RMjvEz08Chv0d0o+DqBhYV0+/CLF/ncvMzfs9Pz3IufG0nv94vD/D35n2bQwnuSS8eD461Mym+Xg3x+ZweMyecXQgRI/EJkisQvRKZI/EJkisQvRKZI/EJkisQvRKb0PJ+/IDXsI4uShXnrbwAe+Pyl6DyYHm+BUW/RCxvib8OZgvvhvzr/WjK2ei7dvhsAxkp823vHdtD4bJD3vlxK1zKoGV8jEHnlF0bmaXy2kV5jUA22vRooowiOFwty8oei2vyEZqM7be905RciUyR+ITJF4hciUyR+ITJF4hciUyR+ITJF4hciU0Kf38yuBfAYgL1Yyzw/4u6PmNlDAL4B4EzroQ+6+9PR87Ha/G5b99qLIvDao6T5yHZlXn1g45csyM8u87mdqnOv/t2VtNfuFb7+wWq8p8CuOs/3rxqv618nDRGKaJ8HzRTKk3zHG1nb0XR+rBXB2oxSiU++HK0DIMd6UfD3zMMmE+3RziKfBoDvuPsLZrYDwPNm9stW7Pvu/rddmYkQoqeE4nf3UwBOtW5Xzew1AFdv98SEENvLpr7zm9kBAJ8G8JvWXfeb2UtmdtTMdiXGHDazGTObKRa7syxRCNE5bYvfzCYB/AzAt919HsAPANwI4CDWPhl8d6Nx7n7E3afdfbo0sfX1zEKI7tKW+M1sCGvC/7G7/xwA3P20uzd9LaPmhwBu275pCiG6TSh+W0tJexTAa+7+vXX371/3sK8AeKX70xNCbBft/Nr/WQBfA/Cymb3Yuu9BAPea2UGs2X/HAXwzfCZ3OLExiqBFN7NuopReD/w4i+wTdpoMbKEwpTcgMnYKNrehrbeKBgCesAtEhxDpJo1SsM8ttLxoGE7tX/4VtBSV3g6svlJwvLGXFlRD753V5+6/xsZOdujpCyEGF63wEyJTJH4hMkXiFyJTJH4hMkXiFyJTJH4hMmWgWnQXiHx+EgxSeoPK3bHPT57fAk+4FJUF76yDNyeYW7Tt8D0JNs896eCVBesjiqhcO8vCDlN2O1ubESxRoCXso/UudP3CJg4WXfmFyBSJX4hMkfiFyBSJX4hMkfiFyBSJX4hMkfiFyBTrVm5wWxszOwPg7XV3XQ7gbM8msDkGdW6DOi9Ac9sq3Zzbde5+RTsP7Kn4P7Zxsxl3n+7bBAiDOrdBnReguW2Vfs1NH/uFyBSJX4hM6bf4j/R5+4xBndugzgvQ3LZKX+bW1+/8Qoj+0e8rvxCiT/RF/GZ2h5n9l5m9YWYP9GMOKczsuJm9bGYvmtlMn+dy1MxmzeyVdfftNrNfmtnrrf83bJPWp7k9ZGYnW/vuRTO7s09zu9bM/tXMfmdmr5rZX7Tu7+u+I/Pqy37r+cd+MysD+G8AXwBwAsBzAO5199/1dCIJzOw4gGl377snbGZ/CmABwGPufmvrvr8BcN7dH26dOHe5+18OyNweArDQ787NrYYy+9d3lgZwN4Cvo4/7jszrHvRhv/Xjyn8bgDfc/U13rwP4CYC7+jCPgcfdnwVw/iN33wXgWOv2MawdPD0nMbeBwN1PufsLrdtVAO93lu7rviPz6gv9EP/VAN5Z9/cJDFbLbwfwCzN73swO93syG7C31TYdAN4DsLefk9mAsHNzL/lIZ+mB2Xdb6XjdbfSD38e53d3/BMCXAHyr9fF2IPG172yDZNe01bm5V2zQWfoD+rnvttrxutv0Q/wnAVy77u9rWvcNBO5+svX/LIAnMHjdh0+/3yS19f9sn+fzAYPUuXmjztIYgH03SB2v+yH+5wDcZGbXm9kwgK8CeKoP8/gYZjbR+iEGZjYB4IsYvO7DTwE41Lp9CMCTfZzLhxiUzs2pztLo874buI7X7t7zfwDuxNov/r8H8Ff9mENiXjcA+I/Wv1f7PTcAj2PtY+Aq1n4buQ/AHgDPAHgdwK8A7B6guf0DgJcBvIQ1oe3v09xux9pH+pcAvNj6d2e/9x2ZV1/2m1b4CZEp+sFPiEyR+IXIFIlfiEyR+IXIFIlfiEyR+IXIFIlfiEyR+IXIlP8FljXDhl7CsPUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(train_generator[0][0][2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(28, 28, 3)"
      ]
     },
     "execution_count": 133,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_generator[0][0][2].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [],
   "source": [
    "### calculating padding\n",
    "height=28\n",
    "width = 28\n",
    "kernel_size = 5 # we\n",
    "padding = int((kernel_size-1)/2)\n",
    "stride=1\n",
    "\n",
    "#output_after_1_convolution\n",
    "out1=(height-kernel_size+(2*padding))/1 + 1 #=28\n",
    "\n",
    "output_pool1 = out1/padding #=14\n",
    "\n",
    "\n",
    "out2=(output_pool1-kernel_size+(2*padding))/1 + 1 #= 14\n",
    "\n",
    "output_pool2 = out2/padding #=7"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "14.0"
      ]
     },
     "execution_count": 136,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "output_pool1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {},
   "outputs": [],
   "source": [
    "#from https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras.layers import Activation, Dropout, Flatten, Dense\n",
    "from keras.optimizers import SGD\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Conv2D(16, (3, 3), input_shape=(28, 28, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Conv2D(32, (3, 3)))\n",
    "model.add(Activation('relu'))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Flatten())  # this converts our 3D feature maps to 1D feature vectors\n",
    "model.add(Dense(64))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(3))\n",
    "model.add(Activation('softmax'))\n",
    "\n",
    "model.compile(loss='sparse_categorical_crossentropy',\n",
    "              optimizer=SGD(lr=0.01),\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "62/62 [==============================] - 9s 139ms/step - loss: 1.0885 - acc: 0.3926 - val_loss: 0.9101 - val_acc: 0.6050\n",
      "Epoch 2/20\n",
      "62/62 [==============================] - 8s 123ms/step - loss: 1.0595 - acc: 0.5066 - val_loss: 0.7951 - val_acc: 0.6744\n",
      "Epoch 3/20\n",
      "62/62 [==============================] - 8s 130ms/step - loss: 1.0079 - acc: 0.5917 - val_loss: 0.6285 - val_acc: 0.7575\n",
      "Epoch 4/20\n",
      "62/62 [==============================] - 9s 152ms/step - loss: 0.9392 - acc: 0.6507 - val_loss: 0.6187 - val_acc: 0.8375\n",
      "Epoch 5/20\n",
      "62/62 [==============================] - 8s 124ms/step - loss: 0.8223 - acc: 0.7255 - val_loss: 0.9222 - val_acc: 0.8359\n",
      "Epoch 6/20\n",
      "62/62 [==============================] - 8s 122ms/step - loss: 0.7155 - acc: 0.7564 - val_loss: 1.0577 - val_acc: 0.8413\n",
      "Epoch 7/20\n",
      "62/62 [==============================] - 8s 123ms/step - loss: 0.6095 - acc: 0.7997 - val_loss: 1.0160 - val_acc: 0.8833\n",
      "Epoch 8/20\n",
      "62/62 [==============================] - 8s 126ms/step - loss: 0.5702 - acc: 0.8051 - val_loss: 1.1218 - val_acc: 0.8575\n",
      "Epoch 9/20\n",
      "62/62 [==============================] - 8s 121ms/step - loss: 0.5210 - acc: 0.8176 - val_loss: 0.9010 - val_acc: 0.8975\n",
      "Epoch 10/20\n",
      "62/62 [==============================] - 8s 128ms/step - loss: 0.4776 - acc: 0.8374 - val_loss: 0.7703 - val_acc: 0.9051\n",
      "Epoch 11/20\n",
      "62/62 [==============================] - 8s 127ms/step - loss: 0.4166 - acc: 0.8588 - val_loss: 0.9371 - val_acc: 0.9050\n",
      "Epoch 12/20\n",
      "62/62 [==============================] - 8s 122ms/step - loss: 0.3896 - acc: 0.8710 - val_loss: 0.7909 - val_acc: 0.9150\n",
      "Epoch 13/20\n",
      "62/62 [==============================] - 8s 122ms/step - loss: 0.3499 - acc: 0.8896 - val_loss: 0.7013 - val_acc: 0.9282\n",
      "Epoch 14/20\n",
      "62/62 [==============================] - 8s 127ms/step - loss: 0.3578 - acc: 0.8837 - val_loss: 0.6779 - val_acc: 0.9237\n",
      "Epoch 15/20\n",
      "62/62 [==============================] - 9s 142ms/step - loss: 0.3236 - acc: 0.8915 - val_loss: 0.5863 - val_acc: 0.9275\n",
      "Epoch 16/20\n",
      "62/62 [==============================] - 10s 164ms/step - loss: 0.3008 - acc: 0.8975 - val_loss: 0.6297 - val_acc: 0.9333\n",
      "Epoch 17/20\n",
      "62/62 [==============================] - 10s 158ms/step - loss: 0.2891 - acc: 0.9041 - val_loss: 0.6211 - val_acc: 0.9413\n",
      "Epoch 18/20\n",
      "62/62 [==============================] - 9s 143ms/step - loss: 0.2740 - acc: 0.9143 - val_loss: 0.4042 - val_acc: 0.9463\n",
      "Epoch 19/20\n",
      "62/62 [==============================] - 11s 171ms/step - loss: 0.2509 - acc: 0.9096 - val_loss: 0.7404 - val_acc: 0.9218\n",
      "Epoch 20/20\n",
      "62/62 [==============================] - 12s 194ms/step - loss: 0.2595 - acc: 0.9099 - val_loss: 0.4705 - val_acc: 0.9513\n"
     ]
    }
   ],
   "source": [
    "model.fit_generator(\n",
    "        train_generator,\n",
    "        steps_per_epoch=2000// batch_size,\n",
    "        epochs=20,\n",
    "        validation_data=test_generator,\n",
    "        validation_steps=800 // batch_size)\n",
    "model.save_weights('first_try.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 313,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.16945671197026968, 0.9546875]"
      ]
     },
     "execution_count": 313,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## calculating accuracy after training\n",
    "\n",
    "model.evaluate_generator(train_generator,steps=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### testing single image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 290,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x14afe1400>"
      ]
     },
     "execution_count": 290,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "test_image=test_generator[0][0][11]\n",
    "plt.imshow(test_image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 291,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "test_image=np.expand_dims(test_image, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 303,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "prediction:\n",
      "2    scissors\n",
      "Name: type, dtype: object\n",
      "correct pred!\n"
     ]
    }
   ],
   "source": [
    "outputs=model.predict(test_image)\n",
    "output=outputs.argmax()\n",
    "pred_dict=pd.DataFrame.from_dict(train_generator.class_indices.items()).rename(columns={0:'type',1:'label'})\n",
    "print(\"prediction:\")\n",
    "print(pred_dict[pred_dict['label']==output]['type'])\n",
    "\n",
    "if output==int(pred_dict[pred_dict['label']==output]['label']):\n",
    "    print(\"correct pred!\")\n",
    "else:\n",
    "    print('better luck nekk time')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "caradvice_ml",
   "language": "python",
   "name": "caradvice_ml"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}