multiclass classification of images project.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import os\n",
    "from pathlib import Path\n",
    "\n",
    "#used for reading images\n",
    "from keras.preprocessing import image\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {},
   "outputs": [],
   "source": [
    "#define path to data set folder\n",
    "\n",
    "p=Path('./Images/')\n",
    "\n",
    "#very useful method of Path class.\n",
    "#This function accepts a regular expression i.e. for what kind of files i want to look into this path\n",
    "dirs=p.glob(\"*\")#this is a generator object , we can iterate over all the values in this object/folder such that it will give paths over all the files/folders in it.\n",
    "\n",
    "labels_dict={\"cat\":0,\"dog\":1,\"horse\":2,\"human\":3}\n",
    "\n",
    "image_data=[]\n",
    "labels=[]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Images\\cats\n",
      "Images\\dogs\n",
      "Images\\horses\n",
      "Images\\humans\n"
     ]
    }
   ],
   "source": [
    "for folder_name in dirs:\n",
    "    print(folder_name)\n",
    "    \n",
    "    #extract labels(for classes) for each folder in dirs\n",
    "    #str()--> path object is converted into string\n",
    "    \n",
    "    #first slash is used beacuase second slah which is for path is special character and in double quoes it loses its meaning\n",
    "    label = str(folder_name).split(\"\\\\\")[-1][:-1]\n",
    "    \n",
    "    \n",
    "    #now we need to iterate ins`ide each folder i.e. over folder items\n",
    "    for img_path in folder_name.glob(\"*.jpg\"):\n",
    "        #print(img_path)#this is image path\n",
    "        #read image\n",
    "        img=image.load_img(img_path,target_size=(32,32))\n",
    "        \n",
    "        img_array=image.img_to_array(img)\n",
    "        image_data.append(img_array)\n",
    "        labels.append(labels_dict[label])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "808\n",
      "808\n"
     ]
    }
   ],
   "source": [
    "print(len(image_data))\n",
    "print(len(labels))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(808, 32, 32, 3)\n",
      "(808,)\n"
     ]
    }
   ],
   "source": [
    "#convert this data into numpy array\n",
    "\n",
    "image_data=np.array(image_data,dtype=\"float32\")/255.0\n",
    "labels=np.array(labels)\n",
    "\n",
    "print(image_data.shape)\n",
    "print(labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {},
   "outputs": [],
   "source": [
    "#random shuffle data\n",
    "import random\n",
    "\n",
    "#note we need to zip both image_data and labels together otherwise if we shuffle only image_data and not their labels with them then, there will be mismatch between two\n",
    "combined=list(zip(image_data,labels))\n",
    "random.shuffle(combined)\n",
    "\n",
    "#unzip\n",
    "image_data[:],labels[:]=zip(*combined)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {},
   "outputs": [],
   "source": [
    "#visulaise this data.\n",
    "def drawImg(img):\n",
    "    import matplotlib.pyplot as plt\n",
    "    plt.imshow(img)\n",
    "    plt.axis(\"off\")\n",
    "    plt.show()\n",
    "    return\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "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": "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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": "iVBORw0KGgoAAAANSUhEUgAAAOcAAADnCAYAAADl9EEgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAWNElEQVR4nO2daYyd51XHz33vfmffV9uzeI3XLDRNILGjtE4ahRIUJVXd0opGKhCphYpKEVWhUDWiCIoqUUhK2iJEoYiKpVkVkobQNHvsjD22x/HMeDwee+6sd+6du68vH/h6/k9kf0hPpf/v4/PXc+97733/95XOec45Ad/3hRBiD+8XfQGEEB2akxCj0JyEGIXmJMQoNCchRgm5xOuiARzKDeB9jbq+7vt4UziI36qB30piYaxFwo6LBIQC+P+qij6YiNRrQah5HVhrZKvqejSMv49MEUrS3YSvv6M3gV+z+6C6fujeY/jNQm1QCofxD+N5+BqDQf27CgSu/rd8v/fyPPy71Ov4t0aXcq2Zj68/9BvqK/LJSYhRaE5CjEJzEmIUmpMQo9CchBiF5iTEKM5UigsQ8RYRkXoN7cGhZkfEW2KRCBb9CpQC4BqrFRyWL9WxdqXkSAEIDr1vDWPNj+vvV3akSxIR/OXnK/g7ziXxd7W1/pa6Pv3DWbin0DIGtaYde6C2/cYjUEsEm9X1RsOR1XOkWVzpDd/HSTrPu/rUjes6Gg1XQhBcw1XvIIR8INCchBiF5iTEKDQnIUahOQkxCs1JiFGcqZRICHvXVSRQ8vSwcdSVL3GkMBp1nAIIhXCaJVfUKz7OlnFYO+wI2dd9/HUd+/RnoVYql6GWXLqirq9OvA73RFCuSkQC4LsXESlXcUrn5Vn9c1clDfeInIBK+/kpqG1ZOQm1XGpaXS9HcAVMpoSrbWoxfH+slfDvefR3vw4171rSIq57H+256h2EkA8EmpMQo9CchBiF5iTEKDQnIUZ5n4PvOHJZrmIt2NAPZvuOw+EhR2TYEZyUVBVHchf0YK3zUPPR+34TahfPn4daML0JtVvGt0Dte++8pq4f+dTn4J7ZWXwYfXl5FWrt8Raobe8aVNeTy4twT7WUh9q3H/kq1F547lmoTUS3quuVN5+HeySUglIghm/xu7/4Hai5DsVfaz+jq4VPTkKMQnMSYhSakxCj0JyEGIXmJMQoNCchRgm4eqzs6WiBYryQg/vyNf2wcdjDaY9wBIenXYeG8440SxJodz34KbhneSEJtVvuvBVqb7/0EtQqNRyW7+5sVddrfhTu2bEX9+d58DMPQe1/nnkSatlqSV0vpjfgnrmT70Jt19YhqNVxBkZWPf1H2yjipkqrl/TD8iIiR37vz6EWCl79Afb/R79XHS2ynPzpb9/LcQyE/DJBcxJiFJqTEKPQnIQYheYkxCg0JyFGcVal3PmZ34FaMVuAWqGwoq6ffPY5uCdSwK+XckybvuBIpbR36639f/b2G3BPLo2rMC5cOA21jVX9M4uI7D14AGo7D96prm/fvxfuOXt8Amr5DK7QSGUzUFtd1atZhrq74B4/jnv3/PTEJNQGejuhll3VK24O3Il7NI3t2Qe1uiNvE/RiUHNVniAp4Bj3fi1Tr/nkJMQoNCchRqE5CTEKzUmIUWhOQoxCcxJilPcZx+AYX92Cw+itnTvU9W2/vx/uKZf1qggRkf96/K+hNrrrCNSW1s6o64EN3IyrReJQCzqaPtW9MNYiuLHWWxNz4PVwWL53aABqyWWc0hnfiRuNvfLWz9X1nVuOwj1Nbe1QG4viZm7rJ3Czrt37rlfX557C90Dk0N1Q2zW4C2oSwNcYDGFroF/G1RTsWuCTkxCj0JyEGIXmJMQoNCchRqE5CTEKzUmIUZyplFiiCWpxx6l91JCr7gg1N7frFSQiIpEYTlOUqrjSojmoVx2Ua7hplTiuMRbB1+g5JnPHHRUJ4536TzA6iNMe9x/7BNSee/anUJt5+R2obW6ACp6fvQD3dHThlE7pPTz12vfw9xiP6PdOOY7vxdzxF6FW+8gDUIs4Mh/XUkXiouGYmI7gk5MQo9CchBiF5iTEKDQnIUahOQkxijNaW2vgg8GRBD7MXdjU+9gkHBG3xSQeg5Bp4GhtSwX3zKm26/1jPnrXx+CeV/4VRyfjcXwovq8f99q5MH8Jv2azPo7hoCOSGA7gyN/hI0egNjWJI6gHq2V13Wvo1ycikiviYoVyEBdNeI4J5+fP6NPDg45eQIMj+jRsEZHZJ74EtcrIzVDbd/QY1CINfayI7yiauBb45CTEKDQnIUahOQkxCs1JiFFoTkKMQnMSYhRnKqURwt5dXsMHzgd79Hb7lQqenfDu0z+GmhfGl9kcaYNao6z3uDnx5hW4JxzA19hwpDeCAX2at4hIqYCnMm/dqqcBGo7xFKuL+ugEEZGh7aNQC3k4vdEB+gGli3iC+SN/9CdQ++Yn9TETIiKrOJMisWaQjnDcO9kVXMjgaAUkw/m3obb4Ez2lIyKSXtLv/dUwHjNx269/Dl8IgE9OQoxCcxJiFJqTEKPQnIQYheYkxCg0JyFGcaZSpKyfvhcRGe7pgFqtolc4tLf3wD3pAg6VRyP4tH8ojP9fBlv1KpjxXcNwz4uLOIReKuEqjMw6HvGw98M3Qa1vAPThCeKeRD955imofeQeXHGzuIQrf770lUfU9VgYT38ulnFOJNTUC7Uux+iNXFpPiyTCUbin4leh1mjg9FEyiX+zkJeF2q8cuUddP/vqK3DP3H9+C2ryyBfUZT45CTEKzUmIUWhOQoxCcxJiFJqTEKPQnIQYxZlKCQVxY63NPA5DD23RUxUbK2twTzSCJ2UXHZ3sG45SkURCTwM0R3GYv9lRiVOu4bB8oYrTA7fdfAvUpk6dU9ffa8LfR3oZV6X81mdx9cPDDz8MNb+mp0Xa+nDVz+VT+uRwEZE7HvoDqJVzuNIls67fIxMv/ADuCRVxhVTfEJ6+HYvidNXUDH7N9Uun1fWqY5THxjJuRIfgk5MQo9CchBiF5iTEKDQnIUahOQkxCs1JiFGcqZSqo3pAenBlRyqph8PLFdy06vb7Pg61/37yeagFHTM59u07oK53DWyDe0ZG8UTp8xdwVcehQ/p7iYicOXsKapmsXvnTG8WVOH/2ja9B7ewEnl4dwZkDGdmqzw0JNPAtsrS0BLV4E65maWrFFU1dA7rWP4abiYVDeAbP8swE1GbeegZq8SBO92Rz+m/mGPYuleDVPwf55CTEKDQnIUahOQkxCs1JiFFoTkKMQnMSYhRnKqW3HzSfEpG5uRmo9Y3tVNfX1nBqxmvg0pOaj5t/JZrwWPTOnm51PRrHc03uPvZ5qC3+5aNQC0ZwA6qNNG4WtW18TF3/myceh3tcjcYWp6ehFhRcZZQr6FUYjuyXjIztwtexgO+PYhmnv0T0KpKQOMbYR3HaZmjvrVAb3Hs71NaTc1Bbi+n31YCPv6zoJK7gQfDJSYhRaE5CjEJzEmIUmpMQo9CchBjFGa1dzeLDv9EYjnimNvWW+q5+Pyf+90Wo1Qo4OhmP4UPPFV8/idwJeguJiKwurkAtEcdfV7WBRwIEQ7gf0F98W2/TX63gURh5Rw+eRBs+VL6+jA/u55b0HkKRFvxdVap4HEN6Mw21SKIPan5V/9xFycM9yXkcJd2zYxBqtTyeiL0eaYGaiJ5ZKAZxsYIjoAzhk5MQo9CchBiF5iTEKDQnIUahOQkxCs1JiFGcqZQaznxIMoXD6P2iH6KOOdIemSo++B5obYbarx7GB5s7h7aq6wkPHwDv7MP9hVz/ZWHH5OUuNL1aRPyGnu4pF3H6KOThQ+DraTxGIBjE6a/WTn3sQq2Miw4ee/xvobZ7n178ICLS1YfTFL6v3wfhME6/7B/Bt/FqCqdZLpfwPRwI4NcMgGtcnl+He8780xNQk299U13mk5MQo9CchBiF5iTEKDQnIUahOQkxCs1JiFGcqZRCGk/j3ba1H2odHXpY3m/g0PVAXyfUQo4+Nq1dOMS+sqZPgO5y9B3yQmWobZaKUGtrwqmgSASnMFCX/rCH9yyuLEJtdBtOBc3O4r4+8Zj+m+V9XAFz4dx5qPWO4YnY3dEdUPOLer+lzTy+jh89+mWojX/sfqh1juGJ464xHyjdszqHx240x/D1I/jkJMQoNCchRqE5CTEKzUmIUWhOQoxCcxJiFGcqJe1om1++fBlqG6lldb29owvuuf2eB6H2w+99F2rhCO6cNACqHxJR3HDrn5/QG26JiHR09kKtAZo+iYgc2rsfaqWynl5aWFmAe2IB/LsUyzgV1BA8ejm9rqfNUkWcAghE8We+bg+uFlpZwo21Xn/5P9T19CSeQh3K4cqZruF9UKs7Gs6hdImIiOfpz7SpV1+Ae8Yd3z18n6veQQj5QKA5CTEKzUmIUWhOQoxCcxJiFJqTEKM4UynreRz+zaVxhUlvt57eGBnVpxaLiFSrePqz72i2BPpjiYhIEExDnpufh3vWV65AzQvjapYwLiKRc+dwtUIxpacV/Cr+zGfOvAu1zZReiSMiEnNM396yW2/IVQ7gVFU9h6t0zs9eglprCM+VOXDoNnW9uOcmuGdtE1fpdOOPLNOzU1CLDuOp3dGwfl+Fy5twj++Ypo7gk5MQo9CchBiF5iTEKDQnIUahOQkxijNaW1q7CDUvikck5Cv6oeGTk2fhnkQMR/Ca4jhi2JLAIx4KFT16trSCo7UdnXgScsox6iBUx/9zC3MXoHbmwqy6nlnFUdfhnQehFh3GDZee/P5fQe2OZv33fPzvHoN7Wvpxv6Lpi7gwIuDjeycKzvR3tuCD6F0xHJLNr+Fp3ruGHNOrg3rxhohIvaTfq/d94Q/hnlIGH/ZH8MlJiFFoTkKMQnMSYhSakxCj0JyEGIXmJMQo7h5CNXxYd+fYGNTWwOHrgodD1xtr+KB0JYz3Pf86Ts8Mdej/Pb09OF2SbMZ9jmYdKZHhYX2KtojI3n24j03Z09MA0Tievj03fw5qfd14ivbxaXxAvFj4d3W96uPKgv7dN0PtunE8JqOYx+mecFRPm12+hHsq+W04tVSP4BRdXx9+NiUCeF+9VFHXe3zckyjfivs+IfjkJMQoNCchRqE5CTEKzUmIUWhOQoxCcxJiFGcqpTURh1opW4JaV5vea+fKOh4VcOW996DW14fTA4ko/n959ZReGRGs5OGe8hKuprjhhhugNr+IU0FH7v801IJVfZRAsoErcXod6YFiEff1qZTwz53J6L+NX8G/8/69u6F2dkqvthERKRZwr51otENdL1fxCI3Nadz3qakN38PhCO4J5Udwv6sO0AMpkl+Be8pVnGZB8MlJiFFoTkKMQnMSYhSakxCj0JyEGIXmJMQozlRKR/cQ1K7fj7WXX9XHBbR1boF77rr3MNSqFVwZ8fTbOGSfCOshb7+E0w21Ik4dlCo4FXTLR++H2r/9+Cmo+SG92dXy5SW4JxLDaZaYhz9bs48/W7GmV8F4IfzdL8/j9Nfo2DjUTs3g1Ec1o6eyUjj7IsUATokE6/j6T07OQW1s1x6o1eP6My3o6WkgEZFYGKfvEHxyEmIUmpMQo9CchBiF5iTEKDQnIUahOQkxijOV0hDciOnkcTytOVLXQ/Yzrz0D91S24oZhoRiuLBjdjSsjOrvb1PVbxq6De370D1+F2mJyHWq779gOtdyaYzZLq95QbPdN+qRpEZEL587j19t2PdQCYHaMiEjyop4W2ffho3DPlGMeSk8dN1Er5/FtF+/Wf88tLbji4/IK1tq7cEqnsIRTUqffPAm1Dx3WG7YVSnqFkYjI8gae3I7gk5MQo9CchBiF5iTEKDQnIUahOQkxijNa278FH/7tjOFJw4ODes+f9Uv4oPRMDh9elgDuv9Lah8cnlFb08QMvnZ6EezIb+HB71cMRt8U1HMkNBvBoBb+uf4+nT+MxE2FwWF5EZDWDD5W3tOrRaxGRekC/FcIt/XDP2Bb83be04mucd5xir6R1rdE/DPfcOuKYsD03A7XRnfg1u9bx6Io33tCj5QcOjcA97/zLd6Emj35ZXeaTkxCj0JyEGIXmJMQoNCchRqE5CTEKzUmIUdyplH6c3ggJntS7VtXTA6Mfwoeo1ydOQy0axweU83ncm+X8kn4dWwf1adIiIuLjw/65VZwCaG3C31WhsAa1ellP3YyP4oP0gQROY7XHcR+b5Wl94riISFuHfiukBE83b23ggoSgj2+te47cCLWz0/oE61qxDvesVrDWPTAKtcEuPOKhJYI/98KinhK8eBofwG/OYA3BJychRqE5CTEKzUmIUWhOQoxCcxJiFJqTEKM4Uyn5Og41t2JJ8hk9vXF5EY8YCEZxWD4RwmmbrDRBrbVlQ12fPI3b8EcC+L06d9wEtVoMp1JuvglP5u5u1tM6M47+PKND+PXOXcSfrb8XV4p84uN/rK4/fSYF96RSOI21cCUHtZk5nFbo7etU1+serhYqYUl2DLVAbWoJj6fI5vH19wzo17i+itNwpS59jws+OQkxCs1JiFFoTkKMQnMSYhSakxCj0JyEGMWZSlnP4lB5Lo9P7YfC+svmSnjK8PgAzs14AZxmmb5wHGrBekVd74ok4Z4vfuNrUHt3Bsfsh0cdoXJH9UbvNr2KpOCo6phZSkNtaRNXrFzf2Q21y0t65cz2YTxW4ecLE1CLxXEzsc1cBmrbR/WGYt0t+DniF/DrzSZxeqMF3KciIps4yyKhmJ7+iuJMlYwcuhWLAD45CTEKzUmIUWhOQoxCcxJiFJqTEKPQnIQYxZlKCdarUKtUcay5UdNTB8ksDnmv5fD/RGscv9fgth1QK2T1hlzZNhzmd/1dtXTj2RpLV3ATLy+OK2dSx/W0SDKFUwA1RxOyjg48v2RgEKeyqkF9nku2gKc1H9p7EGrBOJ4PU63oKS4RkSJ4u8VMEe6J1PBt3AjiKqN8Hv9mCUfVVQbcxsGU3pxMRGTdd8wCAvDJSYhRaE5CjEJzEmIUmpMQo9CchBiF5iTEKO6qlDXc5Eh8rFXRmHUfzyhpb/RCzSvh94rG8EdA2kgYh/KjIdwQKpfFofdaA6cpTk3iUfDdHe3q+sU5PPb8hkO4UiS3hhty1fpx2UQIFBn5MTynxivi1NhYzxDU3prGv2e9rKdMxntwOipbxtoyqLYREVlO4u9qJ2jiJSLyxt9/Xl2PN3Daqffow1BD8MlJiFFoTkKMQnMSYhSakxCj0JyEGCXg+7jnzN0PPwbFcAhHSb0giFr5OEra1IT77LTG8X9IWyceTVCt6Yeer+/H1yE13Cdoah33TTp3CUf+bjywG2oTp/SJ3oEwjv42O7RAAGuSxVHjo7+2T11fWMbfR0bw4fZ4K452bhnQ+wSJiMxeWlbX03k8vbpUakCtsjEJtWIW38OTL/wAapsXp9T1WAxfx2IWR5RzlbT6o/HJSYhRaE5CjEJzEmIUmpMQo9CchBiF5iTEKM5Uyie/8n0sCj7k63m65z3czkXCDq3X0RdHGvj/ZeHKvLpeOP8q3PPAsQegNpvDqZRECz64X/fwPk/0r7hYwl/9ZhofOC8WcEqntQMf6vfL+mH0jfkTcE/7yGGodXThJjzzF3GvnXhYv69iUZwi8ouOAo0QTumcmNDTWCIiueV1qL32/D+q614Q/2bhBk6zFOtlplII+WWC5iTEKDQnIUahOQkxCs1JiFFoTkKM4kylEEJ+cfDJSYhRaE5CjEJzEmIUmpMQo9CchBiF5iTEKP8HdSLeyxbrzfkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAOcAAADnCAYAAADl9EEgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAVzElEQVR4nO2daXDd1XnG37vpLrr3arMlWZJl2bJlY7zbmDVggqmxJ7RACWPTAk7TST90pgNDaUNaT9M0BNoEaGdCJzAtSdoOSxtgStxMajNmicEGG7AxNvIiWbKt1Vruortv/dCv53lnwpe8ZJ7fx/PM0f3f+/8/+s+c55z39dRqNSGE2MP7m74AQogbmpMQo9CchBiF5iTEKDQnIUbxa+KGNavhUu69D/wxnFetlp3jHl8IzpkcuQi1f/7hP0Jt69at+DoqWed4KISvI5eahVoyMwO1W6KLoZb/1QGohcFqeVk8cI5P0Y43LITayeZuqN19pfv6jw0chnNKdfh/ezBQBzVPED92xVLePacagHMKRfd9FhEpF0pQC4Txc+Ct4hSjUnE/33V1+DtXAz6o7X/3Y+cN5ZuTEKPQnIQYheYkxCg0JyFGoTkJMQrNSYhRPNrG943LlkIxX8HzupYvc44Xshk4pzEagZrkcITh9ytpEFjZ9nhwFOFR0qXxS5NQa/CHodZy/DOoddUKznF/BV+HFqWUlOsYWL0Kag0t9c7xtM8dbYiIJKfxfVm84kqoiQ9fY29vr3P8/cPvwDmTw4NQ8wejUMuU8XeTahFKAfAbq8+Vou0/+imjFEK+SNCchBiF5iTEKDQnIUahOQkxCs1JiFHUKOWK7g4oajvwW1vngQ+rwDk+L9bCcbwcXlfDS9TFsvv0gNeL/ydVlIio5lPmlfDph3wKL8uviTY7x6eOfATnFJXr6L71eqgNzuIoSKpV53A+nYZTJpM4GgstWAC1r+3+Br4O8N20+GjoLI6qPjx0CH+U4Hs2PYVPJxWL7nmBehwReSr4+X7r+GlGKYR8kaA5CTEKzUmIUWhOQoxCcxJiFLWGUCwWg1omg1fqJiYmnOOd7U1wTi2IV387FvZAbfjMANRiUXeNmFwuB+ek0/h7+ZV/ZeEg3rjf1u5ekRUR6U+CzdfrluIP8+N6NGfG3L+9iEgB1FQSEakPu1catdV8bTN3aVS5joJ7s7+ISGOL+7cqF/CK9/r166G2b+9eqMXr8TMXCOCaRUVwgCM/h1d/6+uxlxB8cxJiFJqTEKPQnIQYheYkxCg0JyFGoTkJMYoapaTncORQUjbyRgPuCCPW3A7ndC9eAbXWTryJevkVm6A2fOa0czzgxVFE36qVUPMqy+up2QTUtE39L7/wY+d4NIA3+0sB/72CUvbfU1IOCRTchwRGlM3ti5fge3Z+6CzUZuZSUIs3u+O2yxN40/75FP7t773/Pqi99OJPoaYdqECxSE25zzUldkLwzUmIUWhOQoxCcxJiFJqTEKPQnIQYheYkxChqlPLwI38Otaeeegpqnd2LnOPx5lY4p7kZn9yogJotIiJzirag093JORLBJ0hmkkok4sM/l3bSpT6CT+Ps2vUnzvF338anKS6Pj0LNW3G3VRARqSnXL353TDSbxLFHoxIfbbrhJqhdtXod1AbPnnOOd3Z2wjlTwSDUElM4gvErJ4mk6I6WREQKefe99vpwlJKcw7WYEHxzEmIUmpMQo9CchBiF5iTEKDQnIUahOQkxitqOYfOaVVAsg1YHIiJbb9vuHC/m8VJza1sj1Pwe/D9k6jLurtzZ0+UcV7pCSDaHl7zr63FMUVeHl/O1f4GlgrvAl7+GY48fP/8s1B597Emo/csPHoNaW3uHc3xkeAjOSSVxfPToE/izkpk5qKFIKujBp20efPBBqN1z151QW768D2rPP4d/43jQHTsVtQJqUfx8v3PiLNsxEPJFguYkxCg0JyFGoTkJMQrNSYhRaE5CjKKeSmlqbIHa4OAg1NByeBvoeC0iMj09DTXtFEldBPe7SMy6uxNrfTCqVRwtXbx4EWoHDx6E2n1/sAtqsab5zvF33jwA5+za/XWonR1UCmsl8VL/zTuudY5/8OEHcE5YeT5mkziSOvLu21BrancXc/vog/fhnFUrcaGxubK7Y7eISCKLe7YkEziiqwdFyFIZHC1F2CuFkN8eaE5CjEJzEmIUmpMQo9CchBhFXa2dnMT1V2644Qaoeevcq6FxsDIpItLZvQRqly9fhlpQqR+DqFbxCl48HodaSxtuJzF2Ga82nzrtrosjIrJ2k/s3+codvw/naN/5P199BWqxZrzqvWip+5CAp4DbCGzZug1q6zbfDLXv/c2jUHvymR85xyPgmRIRefmFF6H2V998CGqPP/4PULtm0waoTY6OOcdXLXbXzhIRKZVxWwsE35yEGIXmJMQoNCchRqE5CTEKzUmIUWhOQoyiRimxGF6yLyqFeEIh95J9KIL/3kwCRxHZLN6wrW1i9wXcX+/0iU/hnOWr8Cbqzz7B85b14ChoJpGEWjzmrksUCrm7g4uI1Co4Crr9jh1Qi3nvhto0OCRQyOMIIBLANZWe+963oNYaxF27pxPu69iw+So454N9+6C25+G/gNrNX/kdqL3x2hDUfD53PaNAHT40EeTGd0J+e6A5CTEKzUmIUWhOQoxCcxJiFJqTEKOo7Ri+tGktFKdTON7oAJ2t77kHL+UfO3YMaiv6roBaIoG7K9eF3NHN6NAFOKdn2WKoDZ/DdZNaW3HX7u4e/DdDUXccgZbrRXBEJCLiLReh9qOfvwG139u83jk+9Cm+Lyuv3QK1j997C2qFIn529r7+v87xnffvhHM2rr4aat/9Dj4Bk5jCp53Wr10JNV/FfVInX8Pfq1bB78FfHP6E7RgI+SJBcxJiFJqTEKPQnIQYheYkxCg0JyFGUU+laCc+tty0FWq9Sxc6xysVHNtMTExAbc2qtVDT2kLs2OE+oTE1iZfQtRYUk7EpqC3qXQq1RlC+X0Skrs7dTiKdTME5pTJuI1D14hMr586fgtors+7Pu3PrZjjn0vl+qN10++1Q2/vq61B77Dt/6xyfv6ANznntlVeh9u3HfwC1J/Y8ArW6AI6yshX3SR2tcJzHiwulIfjmJMQoNCchRqE5CTEKzUmIUWhOQoxCcxJiFDVKKRTnoDY1PQo1n99d/Gv1etzZ+rbbboPaXAoXmVqwwN0JWUTkxIkTzvGVK/GJg5GREaidP38eatde6+4MLaIX6yqVSs7xaBQXwTr83ntQ6+nphtq3dn8Nai+d/MQ5Xgs2wDkD53DX69YOd5wmIpKYHIfa2wf2O8d9dWE4x6OcrOpXTtUsXYx/q6kp3CeoLuSOGLV7lsvh+AvBNychRqE5CTEKzUmIUWhOQoxCcxJiFJqTEKPovVKijVBbvdpdEEpEZHh4wDmuFa0Kh/FSeTiMe3L0nzkNtaNHjzrH9+zZA+dMzeCeLdu3b4eaWpBL0XK5nHNcO5XS0eVuES8i0jQPn6qZVaKxWMYdV/3Xr3BRsB19y6CWy+H4q62lGWremrtAWSaXh3NGL1yEWlnwSZFUNg211vb5UMum3b9jTTl5EqzHcRqCb05CjEJzEmIUmpMQo9CchBiF5iTEKOpqrXjcm7JFRKpVXPZ/3jz3BvdKBXfDLpfLUPP78WVu2bIFarfccsuvfR29vb1Q83nw/zKvF2va56FV6v5Tn8E5y5bhVdIMWHUVEclO4M3cX92yzTn+8Pf/Hs55Cz8eEsu6N9KLiFx/3TVQ+/dnfugcr3nxindGaUHR1oU34GvPVaGAN6qjuk8VZbW2PoC7uiP45iTEKDQnIUahOQkxCs1JiFFoTkKMQnMSYhQ1SonG8Gbd8alZqHV1uuv6+P34f0F/P24V0KdEB/Pn4w3KxZI7wkB1e0T0TeqhsHsJXUQkGMRL5T975TWo3XXXXc7xFSuXwzlam4zZBN6476/HBwgCPve9efrRb8I5Tz79NNQaFuLN+f+z75dQa0Sb4vM4jtJqXYWj+BlO4kdY3cQeAPGXr4TjwPLneA3yzUmIUWhOQoxCcxJiFJqTEKPQnIQYheYkxChqlFIr4N3+6dQM1IJLepzjWquD1atXQy2rnLTQugkjTS+b767pIyISiUSgls/jGjc7d+6EGrpG7VSEdh1azBKLxaDm87ujA6/gaOn7jz8BtQceuA9qN27A3bJzHvfnJebwM3Df9ruh9tNnn4Xa1ZtxW45cBcdtOXCvY0FcByudwXEPgm9OQoxCcxJiFJqTEKPQnIQYheYkxCg0JyFGUaOUfAbHChX/GJ4HlpqbmprgnAsXLkBtqVJ0SyuepZ0+QWhRytTUFNSmp/FpkCW9+FQNOgWDiqSJ6IWp4vE41LTYCd2z9FwWzhkdxd3NH3roIaglLiehlsy5I7rBN9+Bc8rKO2bbXb8LtZEhfBKqqERZIXAqRYvTtFNLCL45CTEKzUmIUWhOQoxCcxJiFJqTEKPQnIQYRY1SKlKDWls9Xhp+/+A+5/iKtVfBOX3LroRaIpGAWr1StCpY517yHh8fh3Mam3HcUy7hKKK5CUcfZ8+ehZrW9wShtGWRcgXfs3i8AWqplLuTdrWKT1NMTExAza/0NklncVxVLrqvf/fu++Gc/uMnoNaxCBcaa2nCsdOJY+9DrQJiFi2qKisnvBB8cxJiFJqTEKPQnIQYheYkxCg0JyFGUVdrtdUntLonIuIvulemLg2fgXNaW9uglknjzdednZ1Qk6j7f4+2OXx2FtfoDwdxaX9tY/OiRYughjaPd3XhVUZts7/WYVu7n6j2EOriLCLS0tICtZ5u/J1PHDsKtX/7ycvO8f8u42u/f+e9UGsL4bpJAzO40/eKPlzT6thHh5zjtRpeKfd58fUj+OYkxCg0JyFGoTkJMQrNSYhRaE5CjEJzEmIUNUoJBfAyen0Il57PgSVlbwbHFEMXBqEWC+D2A9UajhyqNdBp2IOXvPNKRDSdwXWTevqWQi2XwVFQFrQZ0DaVa928tc7cIyMXoYZaVExO4O88M4Nbchz42UtQSyZxDaFrNmxwjgfC+BloiOIWFB88g7tvn5vBNaHmXXcT1FJJ98b3lnk4tvF4cKdsBN+chBiF5iTEKDQnIUahOQkxCs1JiFFoTkKMokYpHR0dUEun01ALh93L8tUqiDZEZO7iMNTivbgDsXYdfp97+VqLGxobG6GWzeJIRDspopXpRydktNpI2nWEQasAEb2zNfqb2ikd7Xtt2XYr1FpaF0BtAERqsQW4/tF8pT3Fxj0PQu3Qn/0d1LLLlNNOFfcJk9Y2XH9qbBw/3wi+OQkxCs1JiFFoTkKMQnMSYhSakxCj0JyEGEWNUnI5rbsvPgmQAfFGKIqX8tNpvGQ/MtD/ubRgQ7NzvG/lGjhHiyJ8iqbUzhKvH0c3qEv1yePH4JxVG9ZBbezSeaj5IjgmGrnk7ize0oqLeBWVrteHj3wEtXIJdwG/8e47neP/enAvnPPtW3dBLQWKzYmIdHvw49903SaoVarueVr8NTY+BDUE35yEGIXmJMQoNCchRqE5CTEKzUmIUWhOQoyiRilomV9EJBjEvs7l3blCqVT6XJ+VTeJCTFqPEh9Ie86AXhciImUvjohWrr8aahcuuKMIEb2fC+pFsmLFCjhHO1XT3OyOj0REzg7gImqHDr7pHPd7cTG0Wg7HFEpbGYkoheNe/OXrzvGb1q6Fc7QIY8nG9VDb58PfrbeMtYGBAed4X18fnKOdCELwzUmIUWhOQoxCcxJiFJqTEKPQnIQYRV2t9aDlThGZTbjbCIiIBMAKajSE/974mLvDs4jI3BxukbAgiuvR1GruGkK1Yg7OiUXwKt35Ux9DLVPEO99PnjwBtVtv3uoc9ze1wjmeivI/tYy/W6PS0iAAVidTykp5e3s71CrlOajlPfi3Kk+4521sX4w/C7S0EBEZHnKvrIqI3PHai1B76fAbUCvNuetFDQ0NwTnlIr5GBN+chBiF5iTEKDQnIUahOQkxCs1JiFFoTkKMokYp1RzuQDw/7m65ICKSzrs3RB89gqOIri7coXrd2o1Q0zoGj4y6N6NXlYI/wSCOeypzl6AWDuCd3ok03iDePM8dmRx5fz+c07f8CqjNjI1ALRDFNYS27rjNOT52CXfDPnQIHyAYm8SduTeuwzWQ/Gn3vHAR3+fn/vIRqO387l9D7avfwLWH7r3ndqhVxR2LlKu4JUcuhyMuBN+chBiF5iTEKDQnIUahOQkxCs1JiFFoTkKMokYpiQxuxzAxk4BaQ4O7W/PiniVwTiaDd+1fGsXxgN+P69GgNgglUONIRO9QPVfEnblnx3HM0tzRjT+v6P68UyfPwDlXrsRRxFwCt0hY1ot//2LKHfe89M7LcM4f7v4jqC1YiL+zFivk8+7r7z+O2zvUz+AO26cefwZqj/zp16G2/zCOicJt7kgqPTsD55Qr+LQTgm9OQoxCcxJiFJqTEKPQnIQYheYkxCg0JyFGUaMUr2Jd1EZARCSfd0cwLU24VUAkgotP5Ys40qlUcIsHFOlIDRef0qKUluZ5UGtqxN+tXMDxxoGf/8Q5HgDFyURE+j/BBcMSyUmoXen/EtQuTh5xjm++BregqIvgNghe5eHx+fB3Q53Fl193I5zTthhHRBdHh6H2xP63oNbRDJ4dEUlNun9jre1GvoCfYQTfnIQYheYkxCg0JyFGoTkJMQrNSYhRaE5CjOKp1fBu+auXt0NR6+KL4ohYPS4KphXdqsqvv6P//68DnyJBBJT+MNo1ap25tU7U2Yp7id1bwSctvF533CAiMpnAMVHLwkVQ+/J1X3Z/ltLBvObHRc2Uhtjy6aefQO366693jqcyuGdLSbnNFeUETCCEf8eJIRzBfPbxe87x0VHc76e5pQVq/7H3oDNb4puTEKPQnIQYheYkxCg0JyFGoTkJMQrNSYhR1Chl1/ZroahFB0HQdn5uDi/zR6M4ZikoO/pLJXwqJdQACjGl03BOsIbjEi0S0WKWpqYmqKFiV4Ui7q/iC+ETQZU8PlUTDONTJFVxz/NV8AkS7QRPoYRP4kTiMajNJt2/x9Zt2+Cc+S3ufjMiIufODUKtS4mWUmlcwG5e+3zn+P5XXoBzBpTrOHBskFEKIV8kaE5CjEJzEmIUmpMQo9CchBhFrSFULuMdxWhFVkSkCFYatdVOTdPQVgwzs5ed49p/pHC0AWr5PN6Mrq0oT09PQw3VYtJq8Ggr7JksXmXMl/AmcFTDyePB96XmwSvUkQhefc+l8DU2h93z3n3jF3CO1t28olz/vIU9UMsqm+mHTrs3xY8M4I3vWo0sBN+chBiF5iTEKDQnIUahOQkxCs1JiFFoTkKMokYpUSVW8PpwhJFLu2MFfwD/L0imZqFWq+Kl8lAI17GpVt2RgzYnk8eb82vlz9kRW+na3QJqywRCOKqq5vGm+EhI2dyudFdOTrl//0AQb7IPhvD9zOfwxvew8jeTSXd36GgcP4sff4i7Xq9ZswZqrz7/T1CLxnF7DV/c/fw0tOM5tc9Rz4pvTkKMQnMSYhSakxCj0JyEGIXmJMQoNCchRlFrCBFCfnPwzUmIUWhOQoxCcxJiFJqTEKPQnIQYheYkxCj/BxkaB8gtMuBDAAAAAElFTkSuQmCC\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": [
    "for i in range(10):\n",
    "    drawImg(image_data[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "#SVM classifier.\n",
    "#we will feed this data to SVM\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "metadata": {},
   "outputs": [],
   "source": [
    "class SVM:\n",
    "    \n",
    "    def __init__(self,c=1.0):\n",
    "        self.C=c\n",
    "        self.W=0\n",
    "        self.b=0\n",
    "    \n",
    "    def hingeloss(self,W,b,X,Y):\n",
    "        loss=0.0\n",
    "        \n",
    "        loss+=(0.5)*np.dot(W,W.T)\n",
    "        \n",
    "        m=X.shape[0]\n",
    "        for i in range(m):\n",
    "            ti=Y[i]*(np.dot(W,X[i].T)+b)# W.(X.T) is done in order to get a scalar number\n",
    "            loss+=self.C*max(0,(1-ti))\n",
    "        \n",
    "        return loss[0][0]\n",
    "    \n",
    "    def fit(self,X,Y,batch_size=50,learning_rate=0.001,maxItr=500):\n",
    "        \n",
    "        no_of_features=X.shape[1]\n",
    "        no_of_samples=X.shape[0]\n",
    "        \n",
    "        n=learning_rate\n",
    "        c=self.C\n",
    "        \n",
    "        W=np.zeros((1,no_of_features))\n",
    "        bias=0\n",
    "        \n",
    "#         print(\"loss in begining\")\n",
    "#         print(self.hingeloss(W,bias,X,Y))\n",
    "        \n",
    "        #weight and bias update rule\n",
    "        losses=[]\n",
    "        \n",
    "        for i in range(maxItr):\n",
    "            l=self.hingeloss(W,bias,X,Y)\n",
    "            losses.append(l)\n",
    "            ids=np.arange(no_of_samples)\n",
    "            np.random.shuffle(ids)\n",
    "            \n",
    "            #Batch Gradient descent\n",
    "            for batch_start in range(0,no_of_samples,batch_size):\n",
    "                #assume 0 gradient for batch.\n",
    "                grad_wrt_w=0\n",
    "                grad_wrt_b=0\n",
    "                \n",
    "                #iterate over all examples\n",
    "                for j in range(batch_start,batch_start+batch_size):\n",
    "                    if j<no_of_samples:\n",
    "                        i=ids[j]\n",
    "                        ti=Y[i]*(np.dot(W,X[i].T)+bias)\n",
    "                        \n",
    "                        if ti>1:\n",
    "                            grad_wrt_w+=0\n",
    "                            grad_wrt_b+=0\n",
    "                        else:\n",
    "                            grad_wrt_w+=c*Y[i]*X[i]\n",
    "                            grad_wrt_b+=c*Y[i]\n",
    "                            \n",
    "                #grad for the batch is prepared\n",
    "                W=W-learning_rate*W + learning_rate*grad_wrt_w\n",
    "                bias=bias+learning_rate*grad_wrt_b\n",
    "        \n",
    "        self.W=W\n",
    "        self.b=bias\n",
    "        return W,bias,losses\n",
    "                \n",
    "                \n",
    "                \n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "#since it is a binary classifier so we need to convert data for one vs one classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "#build  a dictionary in which class label will be the key and list of data points belonging to that label as value of key\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(808, 32, 32, 3)\n",
      "(808, 3072)\n",
      "(808,)\n"
     ]
    }
   ],
   "source": [
    "#flattening every image into single row.\n",
    "\n",
    "m=image_data.shape[0]\n",
    "print(image_data.shape)\n",
    "image_data=image_data.reshape(m,-1)\n",
    "print(image_data.shape)\n",
    "print(labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4\n"
     ]
    }
   ],
   "source": [
    "no_of_classes=len(np.unique(labels))\n",
    "print(no_of_classes)\n",
    "\n",
    "def classWiseDataDictionary(x,y):\n",
    "    data={}\n",
    "    \n",
    "    for i in range(no_of_classes):\n",
    "        data[i]=[]\n",
    "        \n",
    "    for i in range(m):\n",
    "            data[y[i]].append(x[i])\n",
    "    \n",
    "    #converting everything into numpy array\n",
    "    for k in data.keys():\n",
    "        data[k]=np.array(data[k])\n",
    "    \n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [],
   "source": [
    "data=classWiseDataDictionary(image_data,labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 207,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "202\n",
      "202\n",
      "202\n",
      "202\n"
     ]
    }
   ],
   "source": [
    "#it will basically return the count of examples of class 0\n",
    "print(data[0].shape[0])#it will basically return the count of examples of class 0\n",
    "print(data[1].shape[0])\n",
    "print(data[2].shape[0])\n",
    "print(data[3].shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "#feed this data to svm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 208,
   "metadata": {},
   "outputs": [],
   "source": [
    "def getDataPairForSVM(d1,d2):\n",
    "    #notebook for visual understanding\n",
    "    \n",
    "    l1,l2=d1.shape[0],d2.shape[0]\n",
    "    \n",
    "    samples=l1+l2\n",
    "    features=d1.shape[1]\n",
    "    \n",
    "    data_pair=np.zeros((samples,features))\n",
    "    \n",
    "    data_labels=np.zeros((samples,))\n",
    "    \n",
    "    data_pair[:l1,:]=d1\n",
    "    data_pair[l1:,:]=d2\n",
    "    \n",
    "    data_labels[:l1]=-1\n",
    "    data_labels[l1:]=+1\n",
    "    \n",
    "    return data_pair,data_labels\n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 211,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 3072)\n",
      "-1.1729999999999976\n",
      "(1, 3072)\n",
      "-0.8260000000000005\n"
     ]
    }
   ],
   "source": [
    "mySVM=SVM()\n",
    "\n",
    "\n",
    "def trainSVM(x,y):\n",
    "    svm_classifiers={}\n",
    "    \n",
    "    for i in range(no_of_classes):\n",
    "        svm_classifiers[i]={}\n",
    "        for j in range(i+1,no_of_classes):\n",
    "            xpair,ypair=getDataPairForSVM(data[i],data[j])\n",
    "            wts,biases,losses=mySVM.fit(xpair,ypair)\n",
    "            \n",
    "            svm_classifiers[i][j]=(wts,biases)\n",
    "    return svm_classifiers\n",
    "\n",
    "svm_classifiers=trainSVM(image_data,labels)\n",
    " \n",
    "#parameters for cat an dogs\n",
    "# print(svm_classifiers[0][1])\n",
    "cats_dogs=svm_classifiers[0][1]\n",
    "print(cats_dogs[0].shape)#weights\n",
    "print(cats_dogs[1])#bias\n",
    "#cats and humans\n",
    "cats_humans=svm_classifiers[0][3]\n",
    "print(cats_humans[0].shape)#weights\n",
    "print(cats_humans[1])#bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "metadata": {},
   "outputs": [],
   "source": [
    "#prediction\n",
    "\n",
    "def binaryPredict(x,w,b):\n",
    "    z=np.dot(x,w.T)+b\n",
    "    \n",
    "    if z>=0:\n",
    "        return 1\n",
    "    else:\n",
    "        return -1\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(x):\n",
    "    \n",
    "    count=np.zeros((no_of_classes,))\n",
    "    for i in range(no_of_classes):\n",
    "        for j in range(i+1,no_of_classes):\n",
    "            w,b=svm_classifiers[i][j]\n",
    "            \n",
    "            #take majority prediction\n",
    "            z=binaryPredict(x,w,b)\n",
    "            \n",
    "            #beacuse while building the classifiers we labelled d2 as +1 and d1 as -1\n",
    "            if z==1 :\n",
    "                count[j]+=1\n",
    "                \n",
    "            else:\n",
    "                count[i]+=1\n",
    "        \n",
    "        #print(count)\n",
    "        return np.argmax(count)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 219,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "0\n"
     ]
    }
   ],
   "source": [
    "print(predict(image_data[3]))\n",
    "print(labels[3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}