SBI Mandates Verification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "main_thr_1 = 920.0 # Normal\n",
    "main_thr_2 = 880 # High\n",
    "main_thr_3 = 800.0 # Very High\n",
    "\n",
    "same_upper  = 1050.0\n",
    "same_middle = 850.0\n",
    "same_lower  = 650.0\n",
    "\n",
    "forg_upper  = 1250.0\n",
    "forg_middle = 1120.0\n",
    "forg_lower  = 810.0\n",
    "\n",
    "diff_upper  = 1600.0\n",
    "diff_middle = 1200.0\n",
    "diff_lower  = 1000.0\n",
    "\n",
    "\n",
    "# level 0 = Normal, 1 = High, 2 = Very High\n",
    "def compare_signatures(path1,path2,level):\n",
    "\n",
    "    canvas_size = (952, 1360)\n",
    "    max1 = 0\n",
    "    max2 = 0\n",
    "\n",
    "    tf.reset_default_graph()\n",
    "\n",
    "    # Load the model\n",
    "    model_weight_path = 'models/signet.pkl'\n",
    "    #model = TF_CNNModel(signet, model_weight_path)\n",
    "    model = TF_CNNModel(tf_signet, model_weight_path)\n",
    "\n",
    "    sess = tf.Session()\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    original1 = imread(path1, flatten=1)\n",
    "    processed1 = preprocess_signature(original1, canvas_size)\n",
    "\n",
    "    original2 = imread(path2, flatten=1)\n",
    "    processed2 = preprocess_signature(original2, canvas_size)\n",
    "\n",
    "    feature_vector1 = model.get_feature_vector(sess,processed1)\n",
    "    feature_vector2 = model.get_feature_vector(sess,processed2)\n",
    "    feature_vector1 = feature_vector1.T\n",
    "    feature_vector2 = feature_vector2.T\n",
    "\n",
    "    dist = (abs(feature_vector1**2 - feature_vector2**2))**(0.5)\n",
    "    #print(dist)\n",
    "\n",
    "    for idx, val in enumerate(dist):\n",
    "        if np.isnan(val):\n",
    "            dist[idx] = 0\n",
    "\n",
    "    dist = np.sum(dist)\n",
    "\n",
    "    main_thr = 0.0\n",
    "    decision = -1\n",
    "\n",
    "    if level is 0:\n",
    "        main_thr = main_thr_1\n",
    "    elif level is 1:\n",
    "        main_thr = main_thr_2\n",
    "    elif level is 2:\n",
    "        main_thr = main_thr_3\n",
    "\n",
    "    if(dist<main_thr):\n",
    "        decision = 1\n",
    "    else:\n",
    "        decision = 0\n",
    "\n",
    "    same_per = 0.0\n",
    "    forg_per = 0.0\n",
    "    diff_per = 0.0\n",
    "\n",
    "    # Calculating same_per\n",
    "    if(dist<same_lower):\n",
    "        same_per = 100 - ((dist-0)/(same_lower-0))*5.0\n",
    "    elif(dist<same_middle):\n",
    "        same_per = 95 - ((dist-same_lower)/(same_middle-same_lower))*45\n",
    "    elif(dist<same_upper):\n",
    "        same_per = 50 - ((dist-same_middle)/(same_upper-same_middle))*45\n",
    "    elif(dist>1350):\n",
    "        same_per = 0\n",
    "    elif(dist>same_upper):\n",
    "        same_per = 5 - ((dist-same_upper)/(1350-same_upper))*5\n",
    "\n",
    "    # Calculating forg_per\n",
    "    if((dist<forg_lower)&(dist>=700)):\n",
    "        forg_per = ((dist-700)/(forg_lower-700))*15\n",
    "    elif(dist<700):\n",
    "        forg_per = 0.0\n",
    "    elif(dist<forg_middle):\n",
    "        forg_per = 15 + ((dist-forg_lower)/(forg_middle-forg_lower))*60\n",
    "    elif(dist<forg_upper):\n",
    "        forg_per = 15 + ((dist-forg_middle)/(forg_upper-forg_middle))*60\n",
    "    elif(dist>=2000):\n",
    "        forg_per = 0.0\n",
    "    elif(dist>forg_upper):\n",
    "        forg_per = ((dist-forg_upper)/(2000-forg_upper))*15\n",
    "\n",
    "    # Calculating diff_per\n",
    "    if(dist<=1000):\n",
    "        diff_per = 0.0\n",
    "    elif(dist<diff_lower):\n",
    "        diff_per = ((dist-1000)/(diff_lower-1000))*5.0\n",
    "    elif(dist<diff_middle):\n",
    "        diff_per = 5 + ((dist-diff_lower)/(diff_middle-diff_lower))*45\n",
    "    elif(dist<diff_upper):\n",
    "        diff_per = 50 + ((dist-diff_middle)/(diff_upper-diff_middle))*45\n",
    "    elif(dist>diff_upper):\n",
    "        diff_per = 95 + ((dist-same_upper)/(3000-same_upper))*5\n",
    "\n",
    "    if(dist>=3000):\n",
    "        same_per = 0.0\n",
    "        forg_per = 0.0\n",
    "        diff_per = 100.0\n",
    "\n",
    "    same_per = float(\"{0:.2f}\".format(same_per))\n",
    "    forg_per = float(\"{0:.2f}\".format(forg_per))\n",
    "    diff_per = float(\"{0:.2f}\".format(diff_per))\n",
    "\n",
    "    return dist, decision, same_per, forg_per, diff_per\n"
   ]
  }
 ],
 "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": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}