lectures/cap11_2_cnn_alexnet.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPzzCuzrV7uJqcwmojJjDEw",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "gpuClass": "standard",
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/viniciusrpb/cic0269_natural_language_processing/blob/main/lectures/cap11_2_cnn_alexnet.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "#Capítulo 11 - Redes Neurais Convolucionais\n",
        "\n",
        "## 11.2 - Implementando uma AlexNet para Classificação de Imagens de Documentos\n",
        "\n",
        "A AlexNet é uma arquitetura rede neural convolucional profunda desenvolvida em 2012 para a tradicional competição de classificação e reconhecimento de imagens ImageNet.\n",
        "\n",
        "Você pode consultar o artigo da AlexNet [aqui](https://papers.nips.cc/paper/2012/file/c399862d3b9d6b76c8436e924a68c45b-Paper.pdf).\n",
        "\n",
        "A arquitetura da AlexNet é representada graficamente pela seguinte figura:"
      ],
      "metadata": {
        "id": "cWM11Mu1BE9J"
      }
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iauyYtjtRch-"
      },
      "source": [
        "![An-illustration-of-the-architecture-of-AlexNet-deep-convolutional-neural-network.png](data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAA1IAAAEYCAMAAABhpDcWAAAABGdBTUEAALGPC/xhBQAAAAFzUkdCAK7OHOkAAAMAUExURf///5+fn0hISI2NjVRUVDY2Nmpqav39/fv7+/n5+f///URERK2trevr6/3//93d3Xx8fO3t7XJycj4+PkZGRjg4OEJCQt/f3/f390xMTLW1tYeHh4uLi9vb2/Pz825ubrGxsdnZ2QAAAePj4+fn59fX1+np6bu7u+/v7+Xl5WZmZs/Pz1ZWVuHh4WhoaNHR0VxcXDAwMPv//1hYWDo6Ov/9/sHBwfb19sPDw4ODgzw8PC4uLmRkZEpKSomJifHx8U5OTtXV1ampqWxsbJeXl5WVlf//+zIyMkBAQFJSUqOjo/3+/Kenp83Nzf/9/3BwcMfHx6WlpVBQUNPT0zQ0NJubm6+vr4GBgXp6er29vYWFhcvLyywsLF5eXsnJycXFxVpaWr+/v2BgYJ2dnn5+fpOTkwYDA5mZmbm5uWJiYrOzsw8NDP3//aGhobe3t6urq3Z2dvz++vr9/CoqKvn//5GRkR8fHxISFHh4eHR0dBgZGvz7/SMkJP37+P//+Y+Pj/v7+v/9+/r69igoKCYlJQMCDSEhIv/+8RADA/L8/f//9/b+/xwcHAUKHP/89zkmHPft1AkKCyoXEeDTus7f7BskNBgMC8DW4rmolP/96DM+TXKHm3BYRot7av37/dvLtvf7+fTt4mNUSWJ3jCgzSEg0JlVEOej6/7CahX6FixYVFR4QDe3jzEdab+DYyjVHYLbH1fv06llshN3s83hsZKi7yYyXn/zy4cW4qFppds7Y4ElTX+Pw92VJPJuDabXByXtqVSowOHGBjaGVg4WAfOnx9n6Wp8OvlnZ5fp+zwg8WKVBecJemstDBraixu+rcyLqxpqnA09bh6c7Iv/H199XRxdzi6EA1NKCstdHp9JSmwZqJefj08S4iGce/saukmYuIiouFhcDIzY+KjZqUisfS15eiqnB2dcfN0ezn33Z6eUtDOkhCQPv3+zQ3Q+rj36afmictMGdzd622v+bo4efg2oODhzguJjYyMDI6QEhQVHBscNfNz4N+ha+nnzY4VOsvQo0AACAASURBVHgB7L0JfFPXmTb+YrCvJGMZ8ILMYkAYs5lN+GIwNjabbDAIzL7vO2KxMOuNnWt2TCAh++qQZmtJmjRN0qRpkqZ7m3RN22k7bdp/12/mm1++aTvfTDsz/2m/5z1Xy5V0JV3JckupTogl3XvuWd5znnc77zmXKJMyFMhQIEOBDAUyFMhQIEOBDAUyFMhQIEOBDAUyFMhQIEOBDAUyFMhQIEOBDAUyFMhQIC4FPF6b3W6XPbLstrnsNtlt97o9NpvH5sVnk8vj7pTtbtkl272y221YkttGHo/H5pJkT5PX027nAkn2elGyhzxkc8t2F0q02zs6cNku2yXZsJxuXLS7vB0dra4O2St7XHaP2+1pd3m9dm51ux2dsns8HR2ejvZWl0fqtHWjosSP2gQ5bDYXSNpklzuIPGiB22732Nxumx3k8LhcsuyhJrnVjk/cSlxosjlkQYz2Dre3FV3GcJDNhupdTTK1eEj2eL2y18306fDILR2SHUOVbBX6/DYb2amj3etq6pBbXdxHG7ltHkF7l7vd5erAZLKj30yKDpFb/3g3vtvafd52UNpDne5WubNJbmpqIlsLppyMamIkL5FdanLJkhiYJrfLhanjRts6Ou24Z5BsEmYYOimBci6pwyWBeHI7Uxm9jHrE02SzN1ktmPOAlNcDSAEDaI7N68Ilr8fd4XE1Sd4OlxtkMaiNR4sv21wtVy9dOv2hz+vG/ME8avW4bGgobneg6fynqcPedNpis3sxz9KbbE2tcpfv9GlfkwzQ2trdHjc3w9Vh93bit8fO/XJZTjusXXJTequOKg3TCcnmauu4dN3R5bN1gKPw/BUUZg6DEZHJY2vy+E5brRYZHCvm4EcVbu6C3d3e6kXffdetp61dAJbN47HZ3e7WJtnbItvbPaCJG9zH3dLS9eF1CyAePS/M1cS5wJAZUpLPZ7Fa2zowpd2YEm70r91l78BoMKRsIIDPysmC3qaNi8iSz3ra4TjtA3e6fv20qspSS5NiPW0BjeVYtYAcXVbrdTS2yd1x+bTjukWSrl9XrVbHacV4dnhlCX1TPzzd1iT5Lp8+rXZJLXzlegdQ0h5JKpZS7QQ+BdzYIZ+ABy9GGTze7gK4PPjNTK7J3tHaGvls8DcLOk/bVY9Nptam9nZZYMqDmdTkhoho8Xg6IfwgPySFqMkW3YhgQal+aW/vEqKvS2KO2OFCH2ztHhnVYiztdneHzd3awoW3SLEInWrVBs8JclztxB17qw3ikbkhi0/IZ3C1Fsa3G+xFcBYWEVF8zqDMZC5h5Dyd7Zd9oquyBIp7MPG9tlZw2SYeDA/YGlDm7mw5DTS4XLZutQBqANltPLZIFzohizH+rKW4AakO1n24EigJIoOdxZj4lo4/XRp7BqLbuFBbS3uHaIcrrpiyaU9hpnwo8VNX2/xtsbsMG2WX2rUmW6Quhb/ZfV0+ztl0uVWTKGGPeWV66edvyD6Xi0mPmY8+g6MBI+CtrCPYbS/luWz2Dm8MSIHzMgAl+4XXT/T7Nx/JQBE0P9mGgtxoCvQL2e11QahK1PGrb1ITlICwFqThh+eq7Sc/O/kf/2R3tXo6mjCH7QCR3dPeInugBrq87YBZ53f/o99nbVLaRWRE8/3k6KKX/nR49S+YhXtAjQ4btASgyEas9nntTbLU2TF9Tb/prqZOzLH0Jq60o8VHT/x2yZA3PNAvPSifFWFMZpsHIklmTaSlyXPV1/6nw7+mFjSrOwmAhUZie+HnJz/1MkEXwQRCF1ER0NQOVtzeDthK7id+fuLEmgGj7T6DWZhi9ZL79RNr1hz+WH/y2V6Ycfg3ea2WLvr+6pOHLknQCowL9VKT+4lfoilLvmZr42nzsX+yX/3VgDW48rH3ydAocSlcz4nDH3vZ3mb77hf/99fH2LvweXLO97wXvNGjByx8y/l1slp86mmLz9Xa0eazWFTV5ZHbpMsS1KkPbV94901Z8XYa630wl+xeL7T0F/7oPH7D+X8uuSxWFNDW3i5ZFEubxdfqaWvxtPouqIr7X3N3vUc+DK5xZ1O+alPs37jD+dwNZ43dp5w+rVhBUIvvstVyvcXTpXZC87huufCo89S1Xb9vbYlB6ZQrj3iQydEE1e6nx5gc77VeViwWy4dSU4dPUU6DHC65rQWgsvgufMR545rz83Q17ZAiDwwYCz24f9fx46jgOrdAbW93dygWVbFc9VxWr3a4oCa1fkg/vbjrEbJ2W/V0SW0XHnPuv7H/3GfJJ4YfRs2HPhVfrzZdbrvqa22z2r5/cf/x49c+lk5IKTYQ8fi1+973XHjFuf/4tSufoKuvMN1/SEosSLmpiR7fde25G7uepM7XT+26ccP5afqXXTeOH7/f+QliqRWV7Bbieva/e6/ttb3O48cvXnmgBd097nz4S25f9GS2S/T8xX+k0ySEf2trm8fH33xdUrvrsq/Da7fSj48/C7Ozs52ZXXRipUJucr/29JVP2Nzf/++XvRbOI51u6hTta73cQZdd3qsdPnr8YuMH71ET6wXpTZLtW/c/NdTlPvjVy02CJlJX01UhxFstHdTRau+00Wuv/IJeeubcA8ZES197mBzQ6Z5/7nNvUPvrX73kEw3yXe9oB6WJWn0d9sse+UKT/Qk06MJHzjHjS1/toiSP3O69hGnzX/9Anub/IVUYCE1NV7UWNHXBeu5wtdpc170P3fnVI4+QGmv2mW2X3dVm+7Hz85foJzu+KVvFU1JTJ1Q+VpGuQsPx+a5KdM9zbzDx08lBJHrrz9RB9uuex5yPuOiJHf9GX3D+5joNHAtlKYY26/b46J6LrwJznbYXLv7n+/bOb3/1MtpssT327pvGUsqu0KMPM5F83necn7LTXYc/cdfFJ730J+fX7ZhgEQmuCxsgJb9UMf1PU4r/za5c3fKLnxat+I3L17n3U2T1fPH3TY89d+13mB/w6EU8q/2EameXT7t/6/w3W5tkhzb7k99W/m7iG0QvLMv/8e8W/Lrj9dwfker98Xueb6+2/cvHqQ3WbJqTQu+cu9fua7HZoe188Y+f++rLsuf1yf8352zO+54vfOayS7rwyiOkkJUedH6HutJceURxGjnolXNvkAUavdT+z1+c0vCxNyX66Zaxr/xu3Pu2L3zmWZvvwiufx+hZ5I+e+xGlHVJw2jZd7nwGE+S6TDYLvfDKkf/81Pd89Nv//r9bznz1TemxT19u81z4yG+IHvtc3Y1H6ENvNx0GdqX9if3v2Tquwqpqs/102ZmSV11XO/77wHerG35z+bVnvg7a31Nw77cu3gtIsXsibUmht55y+Tyq9PzFT9PVLph0D33wSdRGcdwTcLnSPbu+Qz77Bfs75973qmC+rRLcVs9fewTiyDBZbG89TE2dbdK3INqUlibJ/viuz5Ltn699naIhZfNI9C1IqbtuHDs19ey5X9hfunPKqT4/cH5K+ckHn7Y7bG895fnCB9f+6zPve1q9MSAFo6XDcuGtdy+3dHksFgkKTVbFqSu/tj9+8fy7W+5/981vOb9Mloc++LSrzd7+9MepBRaaYbtTvijThQ9epNNej3K6/Yk/XFuw4NrnXpY/uuvY7/o4H7Z/xfkJuf35XV8maxvkrfOBNNrGhg1m/2qX+6E70SBJ8qm25z+41qd+11Nv2h5EgyqcD7u+guFsff7il22n29wvfPDxy2pHDHZqWLyZi+wdh5j8pP16U7v6IT1+/7tbCp2fuWR/7NqpogXOj9OPd91ro8edX6N7jn/nAiCldm+Ww/PcRl9xvmo/LSnXr9u+4Cza8kfw8gs/OHWq5AfOf7Q/87lLH9JjV5695+Lnf3voSwQ3sLETwEzXIvIAUn/2+DzXBa9sU3xWgOWz/3R0UZ7dAx9gRGb/T7cEKbXrVU9bl+uh5z7usUIlb4If/jq9feV9RqNRQj0P2zo6r9t5AllbuqyXLzx93/sPPfPUs3BCRD8gsZTyPHH3fffSP9//pOehO+97n2wfeffZCx88SQ7Po3/2uH9841kPXOIxTCB3i0vubMUcuny1sxUi8kHnq+S567mP2++5+OKz9A3nP9Izf26lb5z7BJ1WOgEpi8uTZkg12dBwUt3eDkvHY1cgHx93PmL7qPPJVtsXzt370N2fJNsXrrxMLar9tbtfdLWlewJHUNTdYm9vcT9//ydtbZ2tssX9ypUfke0bzi/LDzo/edn+UecD7eCj9NErL3tfuP8alHGvT057i2zUYrsHXMQCn1ZX5zP3PWujj2JUfgzeJmNu33Xxy0Sv3Hfpwg/eoyeOP0Jt8CVE9CKJnx57k9dKb0OlbmvFQtddFz9JXe6PXHnzoWdgCVy488WmB3Gr885P0j13O53OK9+EmmIwC5OoUJcVCplz/zXnU54fn4ME9HSonm/sumO/0/nuLwhGqy6j7it8v557Lp67du3cD++CzGlzg8nabJLrtbs/TaoxHeCMEPX8H3rnypskNdm9LR0PfcR57c/Ptvs0xV5XPr4KW0q+67lHqMr29Hv00t1PUh09eO5eQKrVwboqvX3jR/Cix4KUzQ2LmyF1CZBqVS48+tTl66X0yrutLF2v3/Xcl90fBU3f+dxl6mhyP/2eDT717nHF8ObjVxPddf+TkrXL52ttRw9KHRf+5UUvVCqb/fH7v+N9BzPqmY+Tz+dzv3PlXooh2qMKTfUCyNHRYrvr/k+2tnVedoFDfdpeZQUrsT947kcyzMlX7e+8+2znMy/S1Sd++fNXbjz8MnUZj2OqDcBzbneX7R7nl11N3tYuYgCV2l+6+/Out69cstJXMBjPPHX5oTuftH/0d5foAsZdgWcu9drgJfaq9Pa1B1yKq11F+W/YS+kbuz5BT3/G8z350c9duuu5z0NPepVOP1Fa9f0/vPumvelq6rWFPWlz0aPvHh28+jdNb5/7kbfL1nGaXnd+coL6/ftB3liMSna1wJb69C8H/88bDCkFYQ0wkyyEWWpvs4eVH/hht9geveP2wau30yuAVEenV+6in86cXHP89y7JEFLClnoNpO1yQ4i8BCbqAGHu7TwPKdXxKEPq3ZdhScWGFFaSWy/8y32XfZK77fKFtx6+fN0h//hdSKnvkPuuuz8v3bX/y+4PoKdKPlb82jxSrN4GupDsp/za3R+3n5a9l3wPQbSetl546ylQiLWbi69CQsCW/Bpdt9BjUE6UtM/fiMbCeY/l+9c+eNFu9ba3uV577vOu6x+2P/OiDC5la78HLXnc+R1u0IdtcGy/7oTDLaKE7v+02WT5+WughNvV5mG1/ntdL33wpAcCytfxlV0PgGH+6PFzD1y4890+tSX77/v097AMmXqlHqwcK+1spZ6Wm6yg+49cFvfj+7/j+cGL5HC/87tn6SP3SW/fd9l1CWtT8uvwqlngLUpLsnV53nrKLsFc/CjU+zav60MbY9lqYzbaFKMSj4DUd2wK2V678R6ddrmUC+5W70N/eNHucxk/Y2+B4odla8X+9hUI4xYomI+DUbXz6J0W7id9dyDy2OMn33X/Ix5LJ3j8hbs/6XFAY/oRZLVShaJc9PapN7GiYQxg+HRsiDLx2d52fsLb0omGvnP+WWqzP/qUixXWDkg/G2y77597wKO0tdLT72HlVxGrrvpWdO+7zeV+68qPPCCI/cLTL1KH98IfPm7H0MrCswOUPXjlWQmIOvc16pJ93asr4dNYOpdtvvZHoZi3IZjrwp3vyar7AvjUg7vutUMd+5r3wp2//xP0asnTcrr1tec+6Uk7pLCm6Gl56On7nqUuuwwJ/ojttO0JTIG3rzyryICU/a7nPvXYU5ceeiV7xYo/Xjz1mWfh103Yr5gZPFgG7KJv7fo0onqg98N2pQ+Zf1x9+uMeBz264Vl6/dzsp5+Exm1zW3D7AZ/SDQCHNQNz960/X3achgnlfNKutOL3t6DvfkiP/u5Sa1OHMffEYruwpapOX7e9deVljwS9z3eJvrLrVe+lGKuWLpUh5bA6YMd8ntqALXrn3Uuen9g+AufI1UgYAlIetqVYSp2GlPK8dOe1H9Lzd7/ILqOXO3917UWX+6POX1/+HsUygQApBO7hkXc/YWv/fs3Lrzs/Zrd92/mI5/GL37G3Q/EjevC54j9fUiwXvIAU+TyWdJE0QN8mKBoPv09Xv/17mN4/pM4fw3CAlCL7PftfpfbHzvzx0/YPL7x97lVqgRYcSYFAIen5RKQEDGOf9/FdD79BFw7mdD4GT2jnY7u+04FZRu337P8a/GzHit6TvN/4X+/bsZjzj7EU+NTbY3e5bar7Qedn/oHafva/6K0rv/aefvTavZ7H3n22o/0r+x8gz6N/hO/8Qw/ShWOPeD6Uu6X4Id7Lo2Jd6JFLtjGP/fCh5x5+0/78B3+GAf+iB1IKCv9rH2S9+wlq/+kIL73wwVOX2ju6AeBwqnR5MKkRHiK1P+r81CXbE2t+ePXp+16G9vekXY0hcUiTUq+SBeGAPG06rg76/aU22zOfa5VjeeC8qvfRh8GOm1wX3nL+xme/a+4vvrDr62T7/v73/EsU+maxlPrWrn+Un7j4MThOnv4MvXTn785f2/Xurz30jYu7rm34w8Pk/tZz5678E/liuFBcHi/+NdF3P3BeO+78zzfdr+zC53tYGHG+6gFDfMTb8tofnL9xWe0P7jru3LXrPy/JUbJS36IUvoPTPHgRS3DO95oeesu5f7/z85ehCtxL9F2o8GBJzs/a6d/3Y2UVObCM0ZMJ3jYvAk98tm9f5IZ8pvW1f+HPR8BXnPe6sFj0Nbvthft3oUH3PHft+DXn7y+70+2tQSwD1tlbO3987Rpq/qr93+907t917us2iOlnO+AleYAw567cK0unLarntf2PeNoQapw6TWBLyR5fy0sf2YXqdn1Twgrzfud9b9gu/OFF+gk9et+zLe7HnC+2Xqd3nMf377rvAaXNHiNoIOk22FxQ/Lp8HrvkfeIjznP7L1551f36fvT2v950d8Za/XRD5nwf7k5fC2KXfnbNuf9+53uX7a/v+nqr4kJMqlGCb/2Z/6SuNllx//Mzzmv7d1154KWnebaf4ZWSyCcQIk6v/fINz0s7f0GXbT/7pufC3U/W9ftUuf1ym+uF1Uc/nL4IsR7fX739kr01NqQwh9CBn/xpx3981ufzIlTjfz4LT2X/X75PnQ/96hOk2r+NUA7F88Ka1YMH9/vUpZgBjZGNM/tbbvLRP/92R+9fSz7X6dd3fOzXEIUv3P4lt+2JIf9g83X+9lOXyP7Er/5jx//M+N8/NF7MM1tTwnwIZmNIQe1+4rf9ev+aEKf17S8OeYM6Okb88kue9rt630sdV7lBNtCr3//3a7tPSnvYIbSG9pbLbbZ/nb9jZ779eudPftZvJwQWfXfn5Sbvt3q/ae986Ve/udzR2dJqp5du/zVizboDKUSLdMjQf+j7P9/xmy9hZfBff9XvN98jpfNn37S32F8/cNnX9O9f/KztdNdPRq/b8cPvyV2tMeK9E9I2OoOn69tfd7VATF5v6fguSv+Szed5/vYd/2ZpdcWUhe144Imfv0+tisd2XX5iZb/eb6DR3x38JkLZY4TKQZVHPR0dckuLcgH1fPN7knT626tXf/N79uuG8gG2LHxwjE9ola6H7n7S04YoPWzMQJgDZLroiAmaQyflIvxRzQjJ4b0diEoW5RJi423pX+MVbRN/EKlvJ0JQv8ArltCg4uM39wn/UDsWpEO5/wLfuph6qAekxAdowUkEacNly35zrW1obo+0C+onYjZFBCvb6QgzRHgfBtMGUuAf4lr5LreJ6dTNxGOsha1gjCVY14S/qA4d1nofGAfkw+1u1qZ/HF1CLUiieputCTwMF3j26bPpvwduiGZIcPZx2C8Sfgdu6bNr39Ej7bYkcYUyEm4gUiY6q3bFgvsilNtila7CPVHFF9Awi0UWAcb4EutR3XUZsWRdXIfNp3KIgixKEYaTonKBssrJTFG6Us1+lS2qRSiUaAV/ylyPvwmiSgQu4l+MtTyztSSRL9AgT7BBGAuPjwmhkVNW/C1OotCksgZawBVxtXDRMEl4iCxBOmhES6rcGJkli4qpzDVYEPKJ6rgOEbEtqzwidg4x5BvpTF3BYBhUL0pHr5WYM91ftTYv+EeARoI2cRqmBI0Vfz2Y7aZ64/H4rv7/70vBZsapI3MrQ4EMBUxQwNXSIWPPUc+a8CbakcmSocAtQgGYml0IlEi3eL5FqJPpRoYCyVOAN4+xbyKTMhTIUCAdFLB5OlpwDEkGUukgZqaMDAWYAjIODchQIkOBDAUyFMhQIEOBDAUyFMhQIEOBDAUyFMhQIEOBDAUyFMhQIEOBDAUyFMhQIEOBDAV0FFBwkCVO6NZdyXzNUCBDgdQpoIEpGG6bekGZJzMUyFAAFFBoUc2aub0VbTNHhiQZCmQo0E0KqJRTOGHTzH1kDW0M6WaRmcczFPh7poCF1pwaSfOyyJqK7mfWAjOb7+95JDJ9v0UoYKF5hQOornoQpW+X5y1Cmkw3MhRIhQIWOjqoaATNq8XR96k8n+gZvJRO266fKGPmfoYCtwYFLDR44IlhVFW0L9ZJ64b9hDdDVSQTSZFUK3L2CFwNW5a5mKHAX5cCgNSIsQ3Lof7hCBvzSUomMw4vSSa7+VZkcmYocPNRwEKzxlPOMCotGpQMpiRaP3xk/sCB+Bc/5edP6JufsdNuvoHPtKinKABITaJRG2bT7kIcUma6Fgutc1bnNopUHDs14lbW/pMZSJkmbCbj3zwFBKQoZy7E1D5STStoCt2+f2/ehLEjR44dOTZOwu289Q0j+E005vH6N0/UTAf+nikAW2o50YiZo2h3LsGPYJIWFpqRU7zelKqo0JIcEidkmiw7ky1Dgb9lCghIOWgJxFT2KlLNChOFTtzeeyqpFkU4yeP9UWjozBGmwPe3TMdM2zMU8FNAQErFrB9F82vJtJRSaM0Mx8z1ZDWhzllo7lzW/DIpQ4G/BwoISOFc6LkTqbRyEIspU0mhfieo32T4HRJjSqL8mUOB1sQ5TVWdyZShwE1NAYVtKXglRk0ZRStrTetngFQ/GrthvZmYC1mlATmIIUzCoXhTkyzTuAwF4lFAQEqBdFrCIRTDzYZQCEjRiY0EWype8XxPtkJMzYaHImHORCVl7mcocPNTQCh+Ml4gM2JFMmtTEkspGjtlsQkMyoqVho0zpSPe/PTKtDBDgQQUwLoUFD/G1LAcEelnbs+8kFIOWr0NYirhipOs0IgNo8zoiAkam7mdocDNT4EApNjVPZv3TamKmbdr+SFVvmITpE/ixSyFhg1LbUfWzU/BTAszFAijQABSkkpzc8hROZwcZmweASnVQTum4r16iXx5Mrwe+VNGmNARw5qW+ZGhwN8iBTRIIbyPxRRCKLJYk0ucBKTwksqyKZsQxpQ4v0rDMmtTicmUyXELUMAvpRRJ8VtTCKEwgSk/pKy0YwtUxcR0UGggAGsCe4mLyuTIUOCmpoBfSsFBIdFQBKTPzzXlRtAgBdFWBqefCQxibWrYRMRaZHb43tSzIdO4NFBAQErT9cSsr8s2tb1XQArSyUI7NprBIMDHIRQW2WtCBKahW5kiMhT4a1FArEtp67WwpiCmEEJhYvlWQAp6nETlEFMmFDolEELRnY5ClKY9JXSudKfBmWf/DimgBSQJ2QHlDE6/qspV5Evo9AtACutZOyabceXBVOMQCtNxuUZDwcpp+pNsyh+T/nozJd6iFNAgJbQxVs4axL6phIgCiBA9gQmOSa6FUCTS52SEJWExOaXTAgOkR7Pyho5Ic5pdZkbIBpqQ+cxQIBEF9JCSrDRgLjkQ6WdNdKhfEFJAyuqNcFAkWJsC4uD+gJjqziYPmdTsOwqy46Si7OwipDg5wm5x3obixZlYqUSzJHM/CQrAlprkjyjitan8DZNoXi4piSyMAKRIUqi8CJMysWCDZINeaWaDVazmy+Q4tqI0L69/GlO5ZV41KYk4SKwWZa5nKBBFAXj8QpDiEIpxsKZwQnoCRS4EKTyzGtZUAinF9Uo0AiEU3TCHZCotvmOVGcstqpsxL6CfjYtwIHzMDJkbGQokRwEofgFISYoCp98UOP1yOZA2bjlBSCnYvlg+ZamJEAoOzcV+/G5ofjKNyVqd60in00+xOGhhdea1JXEHO3MzKQpou3rFI7wMK0IoHEUJQyiCkMJDVrG9Nz4ERQUcQoFIvwRojd18ieqyl9f2ggg1IRNjFxN+RyFL9ZzuAD28uMyvv3sK+KMnAnTgTe0IoShO5O0OQgoPSlS2AdZUoIjYn/DSL9FCKGLniXcHit+KofuKECufRkhBdjY3dkcdjdfizL2/QwpEQIpd3RMhDFYlgIgeUpiUcPqZsUYsNHADxJTPhEQzGgpAKnsxFRxIaOkZPRzzmqRIuYtMNT9mEZkbGQqEKBAJKWzvnblcHD0bymPwTQcpdhSWN5g4008mhFAswfZeM+gzqBOOybrK9TSnUQUKjO6ndE2G7FxUYPF7PVMqIvNQhgI6Cggnuk6P0uJbSwtwCkU8F4UOUigMYmoqbKTE81zbN2UizFbXwtBXSCn46+VCiKlUURkqLPQNQfgFC7th4oVKynzLUIDRAI+fDlJYZxqKw5J2F8r8nvmYFIqE1FjsmzJzAJlKJ7HRMXa5MSvkGxqkqG+BCl0zjclCvWBNpVHypbFtmaL+5igQBSkclsQhFNmH4gYPhUNKwfbeLaYgxYvJKYdQaJBSqLgXOdJIaMTiqnD6OdIp+dLYvExRf2MUiIIU1pkG8vumcEJ6HEUuDFJYcZLLV+DoWYl3B/PqFsIvDGUcr02lfvQsQ2o8WWlOZVylNNkRkBGItbA4I6WSpVsmvzEFxFJvSPHj3RPC6ZdgbSocUhIWTGdMxvZe/4JRcOUpWsPjSD8RQhF9y7iBuqsapLCuBdPHTASU7tG4XyXEIxU0J/Bxxi0h+STTHAAAIABJREFUczNDgSAFIiCF63DgjdDeN2XalhJOvymbIZsob9/ChfsUZeuihc1j9TZaoEKIKbzIg7f3Bq6Y/9QgBTHat1qR07g2xSvciyD5zLckkzNDgZgUiIYUaU4/FlOxnQBhUgowxDP9YE2ptPaOvXv7WZUjWUvG4XxAg61ICk2A+8OMKyOqzX5IwYPSiBAKg7KjnjB5AcemSdWZtSmT1Mpki0+BSFsKuSGm8hHph/dNxRYlEZDS9k3BmqI9tTgqU5Gy1+KgdUkykHO8vXdcEm+y0jU/oPhZqW8lIv3SuJQE2dlcAMkJ0ZmC9NQ1MfM1QwFAanykggaRg8Ch0qJDsWMKIiEFhwSHUDho9LH6PqPJ0lCwYIhVCtloAUJjuReAPQOnXwpTlyGFE6OBWAliCq+2CpTa/U/AvxHWVBpL7H6bMiX8bVLAAFJsWgzcIJx+WO01nmXRkMKOW4RQ0IR961ffNooGTRt+dgjUOwOiwBbCmX6puKwDUgolzKkGCAwKT/USxNSiRphn0Uwg1RIzz/29UsAAUtDicFjSOKpDCEWst/caQArbe7fwKq6F6lezLNpeCN3PAJAIshgIvTK2nRZzKAJSisM6CheaOessZlGRN1jpq0aRGTkVSZnM72QpYAgpWFO8qR2giHVAejSkAMO8mUupTiUqWkmlMq3pEyvIR6EBE7sBKZg7CkIoHLz+xckAttqNJP7KkHy9IPlMK5OxxHcSdSbOGqdnf5H647QwTtPiPJW+W9HaU+LXXfhrj84YfaU7DRWQiphIOHoWp1DwCel9YwV9G0CKQygqaH5Jv8J6ddSyYeNmTou1WCwL90fyeptfSqG7IGjjdtOvbDRBH16ttiKEItFuZl1RYlGbF7Z7MGHixpi7f4naY/FTJgLi1brX7256l7BrNCqxKRANNd2QaV+jnlPSucgpYvzGQ5qE1YvhwhFhKyax0y+GxWIEKSuVNYwv3957URVV7es9P5/deuEFB2qBNTUxtusjkCvqMwQpyLg5lWqs6Rb1YOILWFmzwpqK1d3oAmTZG32xB64YEzAm0nqiBYZNMLyYVO2sa5sAQIwyu19/WMHJ8/ewx8N+GEkpxrrYhsGvRVTDsgd+REMKcRcWmrGRp7kfhrHFKZ8VncLaVABSqAKH7yHSj7f3wrBidz3iibrjW8DgoqBqDqEwN1gIX9+zcnsv/Le9V0+l7dsnGHt40EZ133ZOPVc1l31gOftVowgiCL50Jbqecv3bd+/DiKU+ZFB0FkY2YOWeqsRbSWWwezQ7rOHbF+abHPXA9I/7aaF1kFLRWXhtCmIKLoboe7gSDSlclMW+KZX9fBDLUSOhKwiAhV5p6L3Q5Yr8GoAUX0ekX4GKAoAEiwU6iKpaLNYUXB66OhTsm1IC/EB33fCrSoeuLajo05OpYsG2JauNlGcR7XFtWcWCnqwdZVdcm09VBuPIhufiG7lT+6TcgIo+U7NnsO8r3iQxpLv/okK3n98S3v2KrA0DjWZyeDESTThbG0G5qcd2pmLahxcc+hULUlqkX1VRX2NMGUIK7eLDkhKTCa4MdvrFi8sNtTD0TQcpGfsYi7djIQmMDhmAJ84WzVBDDyf+BqcMdgyb9Pmp1Ms5KPEIJq41Tg68lPzIKIM6GFK7nUuNRyZOeUneUujwALhVo4eTT7/Z7Jxv0DLTNaBrCPU087Y/4yIVWrcmvH7sTy02cf4WFKTK0nCZJNFhrPcY15PK1ViQwkKNcPoVs/CJTsaQksX7pkxgCiEUw2rE7DUqPLo67YoOUgRUHqp0yBzYVD6pjOT+AydMGDkm1pPmrlupudq4u9HPI3j9Rm0p4u2jjd20XbFYaWIOBjuSRgyp7TeW9WzlCl7IN1Co59Gdh5SavmHFSFlNuqusoSNZUPjcYZgAkV2Lrsv4CuLoBpBDT3wLlTWaglR+UZls1TccJ+39RSDF74GHq9tRechQmBhDCoYX1qYwzRNSCmtTeN+UmjBfGEH1kGLLlsWUQkdXlGQfpUVFtVnnD4QzrrCHTfyAdo8z/VRT2ohKt+ds2c2ZYTL2RAKQIIBH4KwOVm/DWs9bZG4fVwL/JO70SBKMBbXgWAOHgeaPO3u21PRjQyu5/otuCIpZaBS6lqyeEiQDILVEMJtQ96ncJKSy80S4UPBJiSFlYPsEK0vyS0wpBeE+cIoIocCbQ6MKjQmpMoRQmImJFWtTSW7vDYMUiHCoCKcISmsnKdOPLLeM7L/4LAJ1o1qaxAVofs2N4LxmClHpaM60bBaTSVSQdFYI87kCOGFPsi0zeE3fYkxpqScdj9qxBgYqEQh1sGLSzIFC8Q5rmtkfCrSMcYihMcW+DAqF4jcgHAbQVgrMSSmGlL5IiYb9ZSClQEyBSdXhRR4GizXGkOLNfDMgpkyofnJg35S+cwm+h0GKX+OBEAowUYnGFK2HgOxdEU6qBKVF34aNIMHpp4YRPDqbuGKho1soF2Iqxv30XIb2PQVv5YqYeAypITlUgJGJFGDpqTZQCviekeLJgdV7amnbCeFyDWRO5hPCTYE1xZhM5rFQ3u5AquivBCnMLlV73xTe3hstFWNBykJ52DdlZlJiuOaacWWEyAhVSITNiisYCuybarRiVdpBOXsBY0sRb89IcYi4TMxUWEjVqqm346i0cwv1LbJGhhzrmpuGr1gsg2s0UpMWkJpIvQotsO+70eGEDZS118FG6J2glIDU4pkjMYBm+E90RTCpsNOVJXD0PTNXugOpv4CUwivXtHHRzILArFS0gHQHxFS07I8FKeAQL5nHDsOEdOF9U4j0S4akekihfIxm7nZyYKdWVimkxaApVYGmJ6zcOAN75FWO9DPRfEBqMsmw5pA5cWeN60t8FeIoH+fIR/ByP6QsbOeaUAgSVxMzh9iII1amwvooIOWgvWvAPJMZwFA9WNriNwQOBSZTIl93IAUpFdYbofil1o1Qh3TfhC2lQQozCmyPd+j6E3RpnEKxsjgAucB1fBpDCjc0p58JV07A+E1CdYmEFJx+RVD8ZkwtJfC8Pv1SPnop1DFYUwXobhjJQ3d13xSGFA2H05FXLHU30voVB83QXBzQGz7gDKne44i2F4OthN9Ka+08zPk41VGNPHnRD6lJDSOjlFJzDWCCgWPjGBIVUyUF8hlCyrx7IqzGHoBUQErBvWmxgu2EBglHhMHlg7WpIMoCFIsJKUg0vL3XBO8ERXm4Iu2EQAVGn5GQwlxuXEn9rs1buHsCLT6WtAfRoA5MUTj9oqVyVFYBKZVKYE1BtkXdTtMFXgEaAV4e3iC/lFJEDGZotNJUp74YPnx4XLTFxpCqVy207XA3NlejkNnoWmrrvYaQMume+AtKKYlde5KqmyDYNzVTi/SL5CSxIYWjZ7M3mcGUdvRsMhSNgBT7+Q8VSbsHnFxycjZNO5qcFqmfObrvOHoWB1sknqeAFO+4HF5UBRT2IKQgzMHLwzVpASlEn1AvHGOl43bpb4Zf8YyM1RSQgs4+ia0pHfGS+ipLDpq4BJAywX+jCjaE1M0npeAvGzgRwa6hGGFIkgE1IiA9QpsHIbUXi0b1lTAH4PQzQSYwYH6lQRLzMQpSEKiwZnhQ2aUQOfDRbTNxBS50U4claVLKQrnz2MHJkzmS65ioLGEWoJWpFBFC4ZdScM1U940QYNyOhKWazwAuy04/a8TwM6RKFASFbYSYQn2pdZ39XysgplJ52hBSN5+UAoCanQ2zmUh+XUaMZ2DfVNhQxYYUZnge3jdlEG0ZNZIWPnqWD/+LumN8IRJSDN9VlVYHVtERl8TBScbPJXMV47yw2oQ1FYDUKqxNMZYRNIwui1DQZKozkVdsB2XtMtg7v5SCf7IXXrECpRDmCMDH/4lkolCzWSCvJ4DvhUtJ4fErsQBS4yGmADfxH1cNqyHUzER1cPA/QihSOozbEFI3n5QCpA5du2PKNGxi9/uRWTbjFAothCI4okyq2JBi7+iMqVDCEmsE2nAlzhcYm2hIYfwgpsSg+JsX1srAg+Y/GZZibSqRmBWQsgDHhSvZ7cZTyW+FCnu7m83QNRhsYwTElN5d7pdSMEPZ6ce9h/sfGjTWE1TVlw7GEqofXBWn70SsTAopZUGtVHNC2JJYcVdx1IiE6GUQI/R03G/cDT7eTjX7gK40Q0jdfFIKOOnVWL/mzCoET2GuoqOAFpgUXuTRGPm+qdiQQhA6TqHACekm2JUiXuRhIqNGzChIMbQPFUEBSmFQdOOj+wpOaxFHzyYKodCkFJa2h1diiwlW7qr6Q1ZaRy7PY1U2ndMadbBnTNfHAKTAAw9wCAV0XhAAAENHMLlN01PX79hfsY0C015XO7L6IYWP8Q0TIKDZi8JX2QRFKxJboqI68CDh9IvAq7iX6I8hpG4mKaURDMTYPXHIxE3nsTKjUQf+YWFNlUZt740HKX8IhYmR5RfEjTI7Ahgt3VKvn+Zw+TceSGsIg4TJW9wc6beOGmIBKSvHIxXuJhy2e6gYjhzaWT01GwvOAGbUAylfwBDw2pROkw5ACnPZUcAxmIo0Yd9iSXEcnLcSKz3MDNOYEOmHYw3CygxACkrMXiGmFKlu+sExkjyq1/w9sBrM1Q5IsZgaan4ChMo1hNTNJKX8kFJp95bSM7NH4GwjhpTgfJj1OCxpZRZ4UahD8RQ/KPYIoVixyYw8Z8DycJlkrEaQEiEUaZxEQAOHUCQMPtOkFPPkfUUOzOpR0xumkTJhLO2biVWydGIK2hQHmqghyReEFHwTB/gVKzSm5kgJyXk1c7fh0PhQRv2IpfxdO8s0bPQDkELvx0+B0w/Q2F18bDzR/C0DimZFe0xi1A1+CENxGDsokuVBhpC6+aQUSLGyDwc/LM+C7q6RAdtHRTgy4snC1kbiSCk8iDP9KjCvYlAydBlDMwH+EKguYSMWyhD+zQhSUDn4fVPgeOF5U/3FfIQsldg3Fb8EASkIBLQcYkq1kFowDc9U9T9Q7GCGlLjz8csP3cX+YbA1OP2CRQYgxX4RqRpiSq6SFsJRwXr6VNj7oWfT8Q0s5iT4HkRisLQgpACJjYj0g2tmTF0RK/sSTT8TsR0p+JTBF+TPn4llt6QNQENI3TxSik+bFN0FpCpwfsRSqlJDPgM4cTHrtxfrSRpXSqEoicbCmgqVYUBM7RLEFBaTw4YrZl5DxQ+7D8W+KY5vSVeCibKwOpHg9EOKIw2HF6lg16VF0zCZD2UfHy5mVroag3LEix/0YioIKSh+WJuC3mellbngTKWbmwtHB0YzbS3QtiRotoBWaABSXhZTM8cKX1SpGPFRg+CwSIK7oQARQqHDq6l2G0Lq5pNS6B4gxa+H1zx+wBn+sXIGVzdcS/pJG19KYZ5he28CPs+kQ5UDN0zCkIQ4YByKGkgpnnCEEAqTJcQpXHcL3spGvBZRdyX6axBS6EIJnH4I2d8KOTGmfF5JXYSSHP1wUlcgi/gMOA6H8z8XgBQW5hVZuGNVhpQq1x2eklua3tpRJdQ6bEkwghRktIUj/eAgojro+jhkJ+vYIDMDH6SAioD0FI4hMYTUzSel/JAqn4KzzQPDB8cflDPM+u04PSgwpKBHAkhJxO+bSsyteEiwvVdNnJPHwBhSHOmXnjWpwDiD9zdXC5dn4Er0ZwhSCNitroLTrZHlskxjVkxP76QGbNjkQEA66C+GwA8pWFf8DSOD+9shpVhUTz4p8kU3OOUrXMdQMe2DRQSlFPPE5RxCoch1kFLQQ+lgwxgTAx8sChw7Ry+BgzfifzGE1M0kpTSwaJASh8ZirQPePgXsGmaOOCG9Ci/yYDPS39MEkMKkXAcxBQonSKxM84s8AuXGzW4IKW4TDktSTZUQt/jgTV6c4oD0ePZZEFLstiqcT8t333Z4mjR/5bRxlSyl0tgaNIupxIEmPGXFT+yXYvMGfZdIhTvWQQewB8cBSJ3caJKYwd4m/oKJAS8S9mYFsgYghQu4h0g/VUEo6GKOgaXZ5/ESpEDGxJ+g3qiGoUkHPBtC6uaRUhAnglgapFRxzJEVaMKIWXG2AsaJlbM5tfqZkghSFuqPfVMIbE9IVIRQQKtInA8FGUKKB7VvkQrgpy1xMP7CSnja4uh+YZBaVaTs2Xhy42A6MHVZDbZ/MybTmuAiwKQOYNwvpcRCB7QsrE1Vncg+NXn5+vrDNWf3QFamO2HfVBjfC0CK65GENUWH6m/k7nbkbDs8ZW6YOpOgKdBqhdMv2RAKQ0jdPFIqHFJQImAHqew9GoiVbcgqdPrw5Emjz+7Trf8kgBTPc7xkXsfYYpJWW0w2NQ1iQArCpHhlOv1c4P2yUolTKOJYg3pIWShrJU9kkAwJkSdphxSAw4fr+Nem/JBifsWRUDKcfofmDG8eWXro6NHxJtlTzPEwvAF7GnwvyCf0kML3bf2Ixh8YvnA6bZrXu68DjTKd4KmF/ws7l01NgFCxhpC6WaUUAgF4Yy4OsaG1YHmIYMOVoacKljWs041WAkixtsiRfqFRCFEj4hv8aydxWFLEVcOfMSDFgVSItDN8JMWLCKpqhokSZ5yDkBLL4fuKqhQV5/Xg5B0wI+hnaW0NXPUQU9hl4Y9ICZNS6H2vRtFN6FDAGG8nSHtiRWV5aNrrIQU+u3zFSKHrgY8iqXGoFt0wmBZWyoEEjr4V74ohpG4mKaURQVP8WCj1q2BbF+7hOxbBUwEa5VePpYVbdBI9EaT8p1CYiTVB/DuvuiTyWoPAMSAFKSUVINIvnukTb3gM7vHCfjFeCGxwy38pCCme7xJl8doUb0bUvBrpRRSqxGCMYF4uCtZJKRaKUhXeBFZVxSOmJjefY3cv/A4fPsyKJ/oqboRBCkcojltDDqsFBLCoiDVMrvNMvRHwZ7JhGF5rvF+GkOoBKYXWYW8Cty2QIGESTFXs6l0P9sO98UMKXSzDKzig8qk0/exRRPZD66gup+apOomeAFIoTdIC0uPRRbunDRcHXSbKawwpPMXvm1JhrSc5mnHqAy0WVcdjCCFIoRSFBhVVsVqWxKSIU7nBLRb7fqcfBgoAE+4JnuL8AyEUkRtvDcroziWL/1gDrYN6SHHLJuGcV3PmsGEb4PXlgPRkqGcIqfRLKZDVcEgNLwb6ZgQpOP1gTYFTQ6Yf2cFv4B1aWZYspDDPEJCeECaAKdam+BSKxPpCTEjBz4wQCjCBuF0NdNnMJzOngl7oQ6wUBilwocJ5PeCVCFXOUUd88qFwjYYghQyI/5IsBXOS1ZxCRZv6hqkgvEjaeIZDCtb3tpMiIN5UUVGZUNpQPnrWl8ToGUIq/VIK+vbaPssWLFsQTPhaMqB/3HktIKUtlEJcYKkXzMICpx9i9CDwHDRyBTZeQkoBUluSkVIsz8sRiAF1JH6CoukQ1lRiJhUTUhy2ADFhzhUfNaRGF9h4aa5GXFGsYQ6HlMohFBAYsXIbVZHUNVjx4kQhQ0hxQHoP1s0NBaJ535TfTtNBSuwUw/besXHYT6KeYskFhyUlpTEaQir9UgoTa0bjopX8Qodg6tWbT6GJ0ycBKagPHJ+vaJDilcN1FdxFHse6giWUkpSC9IHTL7GYAqOlkdnLI/fkGLQ5FqTg9MZL5ndjT0p89CZ1Fzozn+lnLPhZ18NxLgHCcv21sKa8PQUpAAbjw687YfWImZ6m+AkiYYJbqvnAjJ5TPIXGwvum/MYc7+oNHizNQpMD0tGY1PqP53ktORapDWYCKJ/60ZjJnD2BOdwvOMwatal/bvzgeWxrX095Y+Gtkqsm+CGF0eNQct5Lh+Eq2wP3RGU5uydCFEtsS7GFU1Z0YOnarYnT6MUlJVZUlSDFgpSwJ+ZkB8c4QTEmb0u0qhjHjof6HPZcBKQsIoQiVuawJ1P8wQHMeCsXv/8nAlLsvMdhScJuTrFwE49hEojtvZwVFeogBfLT8g1w+qUKaebccPoljP7XtdIQUj0gpWC9LIE3AbT1J1hBYwsSQmrx9uyikUTrs25btaiP4BVgGzjbXOPQ8Bpgww5sqYXJuSfgoLDSrFPLSmpLahOnZRtyEgu0mJBit5ecO2vkpNnpS6OWzx61IfbRs2GQgkNOpdr56dQ8ddNH+8rcLb9BiKkwW4pnOFwEOIUCmnoM/EcVlsIFSEY4/bS1ighI8flhCKGwyClKaXYui3OgzDffEFLpt6UgoGZEHBQN510CSKFtm8r2sH92bH6fnYcAKQwRi6mZmzT6gYVALrOUAqSCfTYjpTAI+yokh1W1JkwO2tSY+GDLOJDCoA7fn5VbmJtbnLb/ckvOInAu2OXwaRgGKREDsK+otCenNIgfPHoWU1Cn+KFhiJ3nZbSerB+iaCC2JAhVIhJSuDp+CgLSY9AqnHLRvzDf4KSHNDBjJWiPG0KqR6TU6qQhBcVvE+UX8XlqNHk3IIVuIQZJpcFYWVSh6qG/vC6lSalgn01BCtN8AQAaTcOoKxItzU286h4LUpqCejB3THk6U95YB940CoU4qrV8IQJSge29xpkNS0j+IuteWHCFXRUJKYRjwppKysBPpXo4/QQ5oiAFjrYXhyUFp0dyhYMjGZ6qG6cUQ0jdJFJKQGoEzn4CirbM1yAllPWhx09aeZAgmfCaK4ZUEku9ghig06oFvP8gLGHlmCcFq2qhJTS8AqwbkAJsvRLtqzWF3jjjFHELq02FMU3+cEiBSAq292KtM72JFfhgiapaRTlgdLBxoyCFEIpiGae5BDP3wBcr5R/BWabsHwm3pcRK9KQVE3i9N0HS90efFV5fWFMmQkL9Y2QIqR6RUskrfgJSo1ZARwekWPHTOA1eLVKcXc2xxQAV4qDZloKHK+hBMCWlEM+6QKcsBmesZFVVbARlYPmvsZSyJkRELCnFhSg0vBAsFCkpv16szCgGeta+XKIY9kEYpLgBoBKcfkEC8aW0J5kmVOI8ZT64JVzxA3OyVs7p4dpR/Oq5EJKYEZGQ4vXecYf9cyelbnPXsDgZmBCJyjCE1M0kpQZWY8sLovS3Q1ELQqp26Lzzt2PTeABSUwt1RySagpTCUgpTLZTwY8Kqcv6NY7KAqQAFu6f4cXkQElkJMckZzSdIqVwjjiBKiIIU7JmDU24f0sNpXkM/dvpFQApzHAdmZO/s4cqHHD15Y16VpLJI1nv8wE8A8uVTBnen+7Pmn+GTYUwOjyGkbh4ptX784VMn19LA+VP6bINyJzoFKZW7nBZX1rNvVLOl5jl7w6nk77IpSGlSKgxSKq1yVlbXHFjO4XN6SG3utpTaBy3N5ICYyyYgFavMaEjhUJXsZUtykk1zk3lg3JJxC7D4FiWl2KE04baNyRSVSt6JS85uqcKxZgaQQgMaGwekUmjgmSXLlgW1loQDZAipm0dKrV87YNbJQzR0wIxZhZMxz7k/MuXl4hAtx8ljeMeABqm+uZVwfrBuhfumIKVQ3z4RUkql0c7CPeuWVRdWzJ8EDYZnLIw1JaD4BeWWEVHjK349I6XMQ0qh/rUmXsNn1DHT1yTaXMjRkFFSCgOSXzkmVmNNl58go0wbcbIGRi1aSuFiYbdeBy7RtHrz56UZQurmkVLaq8kxmRU6EHiRINtSs/ndTYuOskrIHr9eNaUFNVDYhCaXOqRWVeaux/lYB3sv2IhjEjAJEL6Gc/DYPQGp1R1I/ZWllIxldexWgq1hPmm+BzgVzD2FkZhWqGKLVASkmCvRwGw+V6UHEzwlcgWiNMRMCVf8AEVACktj2F2XWlIcdBDHrJvVMwwhdfNIKRHOx1PbSrsR4yc4lab4WSwIxuZtvcKJvoVK6xc4NBPSDKQQPgEpFU4lheZs29lHlCEHwlwl+EBoU26VBfw3oFgassubW0rh2AlASq/mGnZCfxE2EA4+dqimo7DBygsRJBIFKabywOzycFrra0rHd/YmTZ0TG1K5kGCp18OCr5uQunmkVODVNdDQOWxWUAWKH6QU76gD42GloowOwOMnbSkuE3RLHVKHaksbpsOTiLdBwD+ByhDNUnJ41YFa0DO+kMKECb1YNHLs4J74q0upMdCVk4IUFGhBAhABVDbBof+qkGJXzU0NqZtJSmnjGg6p8mKsQQBS0Ek0KQVIwde0BKe6YDqnDqm+hbQOYgoKAttljAypbu2Oqdn7J89bOib+nqkEkOoRj18sxh/tnmAplSyksES6qWJH89L+TAcT6daGVP2tIqXW8WFRPJwRkMrlsw5YcoDz5leOpQNbsHYFZ4/gw6YgJdwTzNtCCZeyqP9MiCl292lrPrgv0Z6GnVsatmm7TKDZqIAcFA3d0hU38JaTUlj7yjk+a0txYe7GIdMdgkxQBsHGsBwOKugpJ27e2pDqruJ300spQMqvxeDlDQVDt8/aC9s7sMpkFlIVApN6SK3KIj7eOSx4BbbB0lqJxszWrsIFgiewcsXHNOk0qQSQ+huUUlCiTzi3kmP5odsn1wiHHbRh7rGIhFCjTBOcj/zXs6V6XPH7e5BSQUgNrR7bbz82P6kBvmkWUnCih4UfYPUXMz8PexNhswUKYym1ObcKckuLSEFc8oDmfGba/gnmh2QCSP0t2lIWOjxlgf+FaIJdYRGwrte0cpU7bySlbmVI/X1JKZw9sc45oErYUTzBk4BUhOK3qhAHOa7DO924mEBi7uuQ2W7jKwqNXDK5uHHY0Wl4CUaElBrPWDRIPRU9ocN9WK3psaVAz5zbFyzEagUkspDNgFRZTm12cc7g5tlQCvx1BmiYkVJ+ghg60W8ejx/WpcTQwVYWG+W51exE1yt+WJeas2wb9iX6rRuFdvBux8BQ+zsa/gHb7BAUvwgpxaF4VM7HmukeFkuYodnLuk/etMF9SopmaJHPqJVDAjVbKjDPwmpLyeOHkjgkkLuha0yg3CSjJ5J1T6BLCLOlmnmDClgYsxcokKz5awdvyc0uFF5x4QdFSCTyQEH+KzrRe1zxu3WkVABSYJnz4EQXft0ISA2EE337Nlp5dhbi8DH/qGz0jh0MqWjqRGiJAAAgAElEQVTdJDAtMFsFpMJziOhWxFjuQEWwlALoiIAUK3xcjmPzVu2LmG44oqAOpwTr5VaospSkFFwBaAJOyxIOSJ08FOX2MKS4OqmKJg4m7MsIwYmviurlCdP8DgufUmblSzLitW9lSNXfKh6/GQFIscePX7LLIIiAFK9L9ZqMMyiqB3CcH1atso4fhQIXgISYA5F/ApDSZxLCBAAZOXM9CooJKRSFPVusCYngd5Q0dNFQTLDSSpyQpi8vWGdqUkq2yOVlogzZqkaW27OQkuW8TVsnSJSzDqdwhpiL1hgILT4xKqAKjtnWuHHdnqEcbXIrQ+oWlFIWml+PYGCeWpGQqh5LvbYAQ2Pw/kowUZnqGop5O1W4DApOcG1i+KWUfqpi5nN0NzbiY5VL8QYeCJdS4gHeXiHeqcytsdCk+im1ObvXZ4+PcexLalJKoUlTiqYOWLk8DzEJYZICtfYwpGhG1pYpB6hmFt72xoexhCdWCzWPH2uFE4bPmpvbkDv58JiluWgp6DFEHCQfIi0HJCUbPRF6OrzuWL8wM3p0qffWkVJs1ggqYuJiHxnHcEVCihU/SCk+3ETT9VQ6umtBHXvVY48LBp5tqUAO8RmEVBk24od2nLHZHbFfSrQC5oMoH/aOUj59xuTKa7XNOKkTiXU27Z7WgJSkFFCz6sbm4WsqKs8Uzp2OTnPHsfqmldyzkCK5fxWOSmPFD8eboT+iejEO2h/wLtFTkEB0US0fv3tJ2VJ2ooOyRxHCzLKMV/L4fvKQknkLoL9sXbVxvqKeHoXUrSOlgpCCjZNzmBdhY0EKcNIGEHOx3+oBK8o1/MUYBbD9OQH3BO8+OFC/saQvDWcphTmxA6dvBuWCAaR4hompxH8Yyfx3ZMPc3WIeCCMIh6hAPdJQl5KUQje2Om/H37zlfccNFzaMXwWTMN3w4ihsQRRtiO5h9z1+cO0NBRHGzYLchZgSZNfXA1oLowoffj2QgTMtywqd2EKTK/csVzk7oo75zLOkIaXhFKt/UDGxfydGN/UNYuj2KKTqbxlbKggpTPRR/rftRSh+finFgNKIDEjtoMN8rlwY0cN++CHllyU4naBmxNoRNKgYQ4M5ULYB4bqB/GwjqIGyAxe1Tx5tVgI5Xl0urZzEB0qrOHtbGrteWO9AFNeQqpQaNAWnQ4lOaDITjZ6+cnMe193DUgqWa9Ed0xlSDprTyA698I6zI1IkiCvmJ0yBKtqa67BY4aZYckdJY0H9mkXT+2P/EuijJq34yct7NW/WDMmAMhDegKhfPQ2pW09KYeisVLNEWFMGkNo+2e9sZ1Lj8MwTRDuK4h5li6PK+VQlgUIoMeOGjewv00FACrPAQv1wrFlgFilBSPHoYob4lTquSsTncFQUZvuYIvZqcIGIXa/dOGD7KH5NGf9OSUqhxL7bavohiBcYDUJqUGF2cf2w5knlYqO8aD3+iE9uDxK+d19KyZa6qt25lDMYkSRSAY43g2OIz5oI0AREUGSvoJ+AFF6MC7G1uZ4boNDRYTR264HBUwunU5VSVQe5lZwthR0G2UcqGgpqc1YuHQkiat3zDxZXYZB6GlL1ZqSU4DIYOByNGWZ/YsYmXJcC70VwXXYe/uqSRMOGgKJRCXMQ9aRy9sRqSClRHOxgHII+cyBXGB9SMlXtPXI7xnIxTgrTz7XwZukUPygrtLo69+wgIaWAGwuN3bAYHgo+AQUzOgipYAlhvcZVi0NFYEHRePD0gftWLdx9dMjJmWfGnS9xaOpRklKKNUkoPNh+sm35FNiPPJtFjeK8fsf4AzP25mYfWlusQQmKqrBZgo1LB6TYDB17duSwwTgFBWeGRU4nzWYFH3GU4jAzJjJItWlz75kLV87vPWPsURx+woxlUUnNttyCCsTdJgspqfpYuTpp0eCJjUVbNeYmAKXp1bqehr4mgtSq8EkeetDMNws2d4BtJszKXAa7I1KBFADFuypw2mzY5IoFKSzzAFKpHDq2WgQHcVcwihbKgZucV1XDlnrFuhSkVKDDMo0/smIkDvpjlSxwMepToUVB9wR4bl0d7d5Aoxu5Hram8J7tQIFsI6gQgrKlqrQur3zkiMXrAyxInT9sybiNU5fVVq4imdelrDSocerUbTU5S9Y006wcv1cjOSnFrI47bKVVy2gjzjwQspE7AMPFLyGrFo88yLYUaKH1EVvw/LwHT3dfSqk71vXKrdBsKUktnjdw0/TR++Y07/THh+A4jr3btvSpz8pt2KgdcYj3x28sbDy1ZWPNxL1D58PjZ3HgzPq+JzdcmzXSgTdAJeXxk8mx7MhqwUytE/I0pirJeWPEIPJeeJGYTkwpbYx7ElIg6J76SLaiNSL8ryT1dxA0X0gpVT/1WErxWeX6a7oneRjzplsA+qFmpVR/zEhH96QUU06BNTUQTBsBDvroCT+kQs2FF/e2SqxQ4tDd0EVdD8RXQAqKnz/cXJjQI28bs7UA0xa3JRrbsHJ03zkHts/fuW6PgBQOaZpeXVRUVFBQXDCV5R8XbTl6cvDt25v3rZ1eJuG17YvROEuQlpjk/vqTk1JM4PI9g2BJAFLrcUZ+ANvcbtyE6GRa8AlJvN/vUO3qZgjkQJLQp25DSqZFE/cO7s+QYjW4eX9ubVZJ/bKKEm33DFo0u/fuhYcGbV26fraKToLFWix1NI33laFpOCGJr47oXVvSO18oxclD6kClOCcM3Qq4eFYfqT+xciuft6NpgtAmoFEwB8Ql/hvXPdEdKYXQndElKgeJxE/YAH5wAb++CQpZ2NtIEX4AxQ/bj2IlOJwqCteDZNn98bguQUrBno2sFqfzLGBhk5KU2sx+O9AMf9jMncjWlBGkNmIsA0miE7PGzxxXCtTjIe5H6F4gDwphSInhwMysOjgq72Q17SlATcgCu63i/IKskgUVk2tqVtHWLFgKOL5h0+LlQweOLO9fBawEl620EiHYtOgJRUTEQQ9iG8hfb7JSSqJ+hVOLDuJYNByosCbkKAk0Ho0UHj+2aGjkvhlbaguKhy0cJWrjDjvEawb0nU42IEnUZKWJszBu7J4pL60SIy2cmag+RFGQV1CM3RRYbIBx6aDegNSYVQumbDxYyohACWYhxaOMf6QWLx+GU5d58PxjJCl54/uuWVBbmFWzVBMBMAU48NI/uokgNbwbih+mw0Gco2oiyVTrnLyUei9B23XZ2VSZoJuhulv+r3ir8LVrU9cPrITipx84QGp+OMj4AeTIck7eTIOX+NlLsJREBzhj0ZUhxTNf8YHWFpq9IR9qu4GU0kMKatsayts2ZS3qQ/sAcX33tNr9kNKkFN7Vh21BtcvpYKMGKZC/Zl6gJ6z4WVCKCBfwP40m8TfZYrXiEGj29WqR6DgEkGUcErvCAvwmWSklS3k4UbcIkAKlp5QZjYXw+DGkxACUbmreUbxO5OMJqdB8nXuFW5MspBBexK47QMrCkKoXzEmyOCwtGrI4fAR0xWEF+BAV8Au/pvFSLzgAIEVLC3uPYN7EK3XmN8pz6yH8ydq4OB8vB1A1WooKNNEkDx00S8wJJr4YAwCLtQHQIa6U6g6kwMCmNcyfx2n+/HnaF/Er4s/8nTu3j3M6r9VMPcHsgJutJUT0nClYVo9Uwn/qS0q0T/5Swtf4zx1O566ChihIHS45EF3j/F4TUc/eihOBSarVk/hM9CCkWCQww5Jo4jBeMYpW/CIgtQOOgvk1UENkuRR2HKsm4QnwDEop5ozUvxwo2xOC1F685gwcENIG9GRICScBpnCALYry8KDgogCQCJvF/Fk7Fa/sxmTi9RR/rUlKKS+eGzvyZA4f3mlhMcX98Dff/wVtFef4cS2Y3OKmxjggFMrG0/y9/FCIkyQLKQ5yhMYXgFStgz1MTCf27yFpsomvBISEgFQWL0RbGFI4CBHQAAXR0mQgJZEjD7Fk1uLp/HIA/84SUSPYF8ukYOUgzaqaA4vztM4zhcIgFeo7P0HdOnsCXZp2tqamZhzS3Inig78i4aLuZ824mpz6XU7nmUY4avUNwPTKnnVoYXNz8yL8b5QWLtp+yunMHbwiClInZg7TVaFV66/nSHVqkILmocjlqw458JK2zYv2NEBMJYCU2NyBLvGUVmhenyoeGj3TYCJrkApGomPKwh8A94QAH+7W4F0XoApmDZzogFQ0KnmgA4l9cWxLqX1rK2f192fmMkWGZKUUJvPg8+eH8oEziEraAGtKoIPnsQYpPhcetpRQe7hjIlJBCEUsENHmhtwNDQcWj0HveVGYzw9NFlLcbnTcD6mtJRwxGaQV3wt0PPCJplloayGDX0gpVKqhkDOYl1K8/JDVuJ0cxZtp/EyIqWBF6LjWKHj9NA6Dw+BObqtesWzY4EHl4HgyVeB1i/DtW8XRmPoZjSe7BymFDi7z9zQvZLYG+q77VKjGueEo3Q6FTHcV9ZcVx1f8FGp2Fi2UJ0R7/OYeDS9LlCvROOeRnTQkFcVvOsrDLFldfL4/0cE+p/atWQIBxIeO+WmG4eJIdEipIPX9+6VEBsyljcVj2CEeTmPMF7zfF1PU33OeqwqtbdSKwd29OwFdf/M1SPlzGn6wn6Cuev3w4qKVdbqW+PMmJ6UgCS3KmP6r+7CUQt8na04/zMyASMA3PsAZihjePKIpXixDhSCFImYpz8+5bVxtY8khbRayDprC2RPoP9wTwpYSkDLsuHaRpRVsqE0lY2/Ph9cT7gkIqQBpk4EUfINLR249spgalxJtPIzXOsWuVbCZMSPWzhpXuRPLZwJSYH3sI8DqhSaz/U8DUuyeEAwpOOSxS466wyckwUy0wotZ3AvrCg6rVVWwcsInSPEfhwOsC3+raPrM3nkyv/cpAlJ5BbPxFnAlv3A9OaqQE2q1o6rKUQUFGx/47F8zv0qmUVhMDdEN7RDuCeGGYfUokMA2j/TuL9Hqk+H1mFH8ZgBSGC6rY/YGOFPrqOLAWLw5UZyQlBBSTBjmlbSkAc4jbRNdgFgse6D4hSDFN8IgVbMz0FrNlgrraKCY4Ce/1Lf8yIbCRTjwDz7jiJSclGK6cVum0D4ofiqth5jCuhcnjkXgslmdGoRN/fjBihc+9FIYMJy/jcaMaD6xdfPS0XPWQ5olK6VEByw01zyk0Gll/o2CiqEapMI4mHkpBfsISlLRdAGpSRBTfrYWQVDxE9PCr11V1WGOEE1dREO3lmEWbq1Xw98NLSCF9XKo1Mx54o+lQV0MKZV1OZlWrARLR2JVn5Pwv4L+fPQJ0oixhH0+kUuw2lIvRmHgmaVQmZBP9TcdaOHHZKkOnhxFePzCWme8LgXXIFQXh4CUPrsZSK3WpBTUnzN5mBm0YD7VDCPZ0JYKlu2XUtxUTD7Yx+vOLI/gd7I3Qkpx3jBI7Q06WkxBClrX7Bs7hTM56JXgMkVKTkrh+ZFLdi9sgC3VByOHN46twTTRkp/FKsrAgfOL1k/bOnpfvgAUILa0z9SKBfVQmybWqezxQyz+gMJjzhsFvRFRHJRSHHUncKhZ+/5ijT+gxQwGUdD/xFKK5FEnio8cgfIAxU+TUqFCMZWKytEV5jVsXQVHKpTF/w3Cfv2yO1YLxc/KbwTV+YQiM8M5o9m6sNgsiqpQRTMdys4t2Xagd2FdkGLiKQEprl1z1MauP7IO/2+GFGbfpC0LRlfvJmVUTtbhMlJWbh1c2BsMdM+WBXtx+vHyvYWrx+C8BnDBdVFSiqMn8tcd3ZS9lOSRR0/sUXD0Vtn2HWB3tH7GjoV1OEKld7/RWJeKllJDwvuidQj6NYz8SOiahZRwFS8/019AajfeI58PW2p5oB5W/MbSSr17Qg8pDLF1j2XhMXiJ9ORiKSUUP/1FhUbjTc1i9mI26RQ/ONF54hokqBJ4gk0ctqUQ44enozWL5KQUqqrqtSRnXpV2eCfeOFbZb8DcnJptG6dOhNGAVoBPzJ2yYX9jZWNuQV+Icb5EQ2fN37194aK+q6aDiYh39fZfNLlgySZWd6AjaYeOiaxaA/VyzaBnuAQpVQINDrZkTEihEC5NosVTp0xeOp2zw6bC5o5IKcWQ0pEmmkZaEwC3srVDsgaylHLgxdUD0V8mrt8vwc4gvKWPf3LFLKHBCfkPyhYHOKujhvceN/N41jzmI0FxJFPxcNp8dBrRmJ0D+uH0tVj1i5Ki/wjFjyYc27t9y65FNPtsTfOy4ioqum3IvFPraP3Z1b0GHKTFx4Y11/ZRQQBDSBUMzD9fcbL6/HqadL5+zW3rKP9M7txtp6bTnvNzDy9bT0vPVxw+tnvsCnOQAn/igISUIDWNBwmNXH5Ek1LziCYOoLFhkCqKBymM5LDsgaNPzeEhDSaWUngnlc4nhlsI/S72B5QBUkcDIIQikYWBDT6r/wI/Fy+OCMuhFDF+POGjU3JSKvC8eAMWTygq3Lh6x7pZt+/c3Vyn1SBhXbVvLnQYNNqA50NKbaGlAxoKd8IBL7rtV/ywRLvu5ECiQdu24SOMyQRq1X2CS6zA2i7cDoaQ4oGBKsrggXzvjWje6fw6Bow0r0uFUZsGriijSesOjyTaV7NxDdwO+tu6KlnCS1QwnwqWss677aRopKA9owCICz3IQOM7smPo+M1rB81ZuEI7x0mi0YUVU7ZVFubBCa8lSKlB/cet2EuUf3jIiQYcZZ0cpjRIHa6EYrd/Ie2tx37TKasoG+9en59NC8/mMZWXbSSqO78HtDA8e6KgbC+WthZeG0+FeGz5bRPy7hhCVJ9DawqZRVL1DDiLsxdjXSpspgnFL9Rnf3+0j9SWemcISEH5mb2hlL3nFb0IIRQjx4RHT8SDFLvAVp7fvmd7OKQ0KRVOWEipXM23BqqES6kYkJLFai7MNLZn0njaLA6PYec9XtfDg2WBK9FPVc3q5qnMB9g2seYTYMW8xYjtLcCbFb8BM2umQbYiKl5MO01KSXRo7h3g1cN3bt2L5euwwQsbLf4B84B2ZiEuIoaUQitUMGW2KbgOeNrErl6GVA4Yoa48tqXGUt9xx+BzWDVv9NRlMSGFIRg+9NBt0xhSqH7xFKDPUTVmbP4k4fVBd2ntjhlrBswdt63+KOpGvRKNyC3Oql1WMflYLmwvBe+O2zp185Zjx2Zw1LKWAKnhxKIbeqRE9RwTEr/vusbzV4aUg2pPUik1rKSCMxUlfY7Np6KVRH1nUl2fUyXzq5QNDRUlFTegNOA/gzfKF+fDiaKMOrtYPbbg5NwB19aXwwlDaybT5tuK+y2m/Dv6nMwZd/5QccRWzfiQSl1KYWiLj2f1ok0Ljk1ZM4YmnnQ0hnn84kMKc3NocZ+ywKwU1PLbUmGTCnNhawOUYY2GZqQUT6eyw0eBdTZc40IquXP8oE/xOg88fgJS1Zt51whvG2HPBZKEg+/3Fcs8oYOQ0volZoqFbt+4fgKyYRVam1Z+KQV8SpWIQ8XneGjS8acVTlmTS4sGwSgzlFIQEqiSEcyQgvvLoZ3jBzIaKH7ZZURViExXQd9pU/qjCaK9kX9Uas7qk8tOdEAKmBp3W2F1dUFjcTHOymU4gBP23Xu437oh83st3AQOwqXI1vIxpezln7wPI8G23/ZTuQuaMZIasTiLgNTgjQj8zGs+WQ91MhVILZuL5m/oRdXj8jctHlFHBbsBqQYi6/ohd+ygI/0Gblo/FF47Y0g15lfOJ3nE2cWWO/rtW9Q8qGoCIrPpMETbiKOFx7aW3TFj38JF05bDiR5GmR6BFNQL7Og9OH3fbCobtHXP9Coa1TC+VnsNBQYUwS6s+G2D7swaNodwC1sKckXMN/wFppTVZxeyDsXKuGgyyly4BT/0Y4pftbeDahiCCClVa2RLcVzAiLkrahBKyuUGlnr1Jfq/a7YUVy80F38TDDLqL6GFQkphIlVrwcNsuIgcKEBIKZ4YYm5w/fiqqNp9LKMUlbKXkGvUuhuwpSRH9jRYKaM3NszD0MdNwIyDdteDIBG2FANTnJo4YUifsX5TBtyFI/aZTojxr2ZxEkrsnigjKq+ejqqH10zZjqkfuqv/xkXXoRArQwqd7F2/dPr68aPyR5YzZjCkQZTgKT8u8AzfUdjjBy5gOZS14jiWD4JvihDls5QaPBl9KVvTuKCOuYW4bPIPpFS9g+YdG0vDnc3wdo0hGltH1SuJDk2hCWOJBjTSgBXwBI6EMDWGVEHeZDDIA9cmUe4SoF6lCTMBqRMbaQw052V7qWEHU9Kse0K0O8XNHdMwZ/zTQps9cKlQzsYto5jCnLCSwe6JbYHfYJkq9evHGgCTWktYq5i+YRiGnREnHjSEFKbiilKenFGQMlD8UMzycTNzZqNiLtEPKX+rAhVrnzh6KQu6Dp/8DH1OVYF1w3yRT63qg8aEIBW6zcMWfAsi+xkwf4RHeQxW78T0qh2CCkOVBKSULDtWbIXcWT+rHpHeofuhooPfAP8mSa6CmIqCFNdIVasmV25bVeU3PvGY/xw/LqD4AOZMKGkeP8oTkJo+pASKV8yqmfoISPJDavAa9DQsMaPUUrAI4BdYA6Tw5g4EseTMqhqwhqr0C2NEhcOJZm3kemE9rElF8VPJsWDK1Irzi6h08pmKBQ2TqHg79Nhs6nWkT8WRvjRm2cypJfBQxVD8GgeOuGPBgMo7lmKfRNawqcvyRp4BpE5W0KyCJTUbBtHas/XDFuxdnpwtlaLiB0ghthoJypD4QED6kTPbtx7cs3X69K2TsL1Nk1L5+QO1hABPQArZ68rL8/Ly8CL3OglaSdlCrE3xC0PFCGmQ0k06XGXeX3IUM409ZGG2FCKsIxJGfvGWDbDxYfaYgBQMH87ltzA4sCOiPIOf/OpTEbMQkFL6PAJSomLmNFj6BYuURwyuhnIhmMoe8AYdKw6sS2FGYdGH4/ZGHknw2gGUo5biqDcsf0VKKUBqVL/qxt4juUfBrkCjE1IKk+pQpd6PgCbls+I3Bm485k0jp/CLqvW9CfvOXESDlEKzOP6MlQ9kj/0EboGjaAFJQvMbMTM/jMTw8KyClNpCljFY/qnntfM4pYW1RvxgWwoOROvwA/3z2em6eWXzCEkakYeTpPCis+m9mnmDBG1duWggS2dDW6pxFC1fM3j2WpzNM3bImnWjLaWD8ojGb6ZRQ/qtxvovDRzcb92mfAQkhbWspxQ/roUJjf9Z9FtpjTNrGSINs7IaT0Dxg5TaXVNam13JqQiGrZBSWB7ZUJSNVHSEgxghs6A/wiGjEdNQSoG1jobGBEyFK35rDSBFas64kWCoKqx8JgLOZVqBVYaQZNQNDN4oX5K3pmJLL8o7OW5YTm8ryk+cYDbwXmUjKcXDpj97gltQ16u+oWYtZgwghXlV35t9dYEUhJSjbMM+pWp5nnzwzFAAK06SaGVWycZ8CU6/tX6Pn3+wwXCGTKk5aOUgPsx3/1WU5ocUWtO4ENQOJnZPlJMjf8Ugxbo8j/admRAmxIL5tC+840YNSClWkvwpUE/gt+4Tg4BfCEiy8ICAf86F2SOGJZApd5Da58ix+vzh2XMLqyeE4S2QJc4nQ4pjWniY8SG6BhVUeKcEm4Qmqa3w8y9jSPF+KST4dERHkFE4S7XuIZQUX6D4GURPGK1LibZ2R/ETBWgUBYMEX6ybXQXXNSI6qqo0KbV7L5WNLUMqH1uGvkFKqThBYvaIUSOQZk+AowxbAeAUztMYHlrvt6X04wRyqVQP5znvONK7JwKQQmYNP9wgqQrqOpcrWgdylBayc1b7Ff4Xit+yxUvWDjqziFbNGZ2L8wjNQArvttqm+bn8tpS+UE3x8zNvhEOtH9CQtRsr934qqfC0wOEenFV+SMHjV7yi8jCdqF4AWyqOpOCqpLVb8yfW0/zcgHuC2ZkggUST8kEKDAGuBAgQghTQtKgamYPtZUiVcdXZM2hAdT0CENSgbAtm0n0JQApSSgcpXQb9V3bkg6HsHqRBipVAnFs/c5Q2g7WccE/so6Xjl6+vq1q6ag/WgUOMQF9UzO9wRLKUwsTHngPGlOg6d59XEVREJ7HmidAkoRjEghROwlbhP5EUhDOBEaNA7DJT+JegIwKbxBbEsCnUI1IKxBfCSYyR+IYGMKK1BFsKih8gFbjAaGHFLzC9kE1YG/yIKEo854eU118Kf7Cdg/iw7DqoBdFSSptLGC7WC7UimDcGfgjzC4Ih2ApdwWgLK35oFi+qWxsPxhBm+kfQWoXU6l6sosRQ/LCeAUaJxiDn0IKcTVBAxeziYiBI6vXWVMCWoroJ5RPKqGr5UpwTY9jWYCNQMA6OqpYd2WsXZzngf+O4C3EIr+gz9kFqdPBTA/QLSCm0QypkMRUQYBqkxkwoG1lOpbOXlvGUDtZj8AWKXy7WpTAKCSGFOQ2xsPRkQe4g6nMIXILHCZxxYg4oF6gf4w73BLMx0VgQJzhsBrUbXBJSih8SxWsf+mz+mSVux5BS2stwOAf+aZxJ0D9QGBNEO3sibFh6AFJbeRVcsAAra+1gEtjMwAjHNbQGHxqk4PFjqDPTxAwDpNBEYXnhSpCPMra0BrPZtJDVqiBh/OSA0w8n1TKkdEu9LKU4J0wyLtaPaG5A8HFGzOgGPnLVIAn3RJXUH7sKS2lRNuLiw540eAKXeGI0V4q+x5JSeLUqeszKcBX0ckEY0SL8AV8NLAhw8QFIaQ3m2tkhZlyx/yrIuDv3DBaxdldsqnXIvEu1dJR4mytC5US1EY+HIIXos0MFOHcjQF5N8RPViaqZK2ktiSgiUDX4zibAIzGkmFmW7y5sGLAeXfRv7uC+0agNI8KAC0hhArFKz0a5cbWxr/oh5c+gD8mPfIaLNpRSZt4vxYrfX8SWgtgRSyCAEiMG2991NMHQ+KVU6GIAUoGB0w1f8Gs0pFR7r4pxU/fQvsoqVBPmnvBDiqeV0jxOhNaAcsyadAnPLBgCYBkkaGlZslRauxr8gbJ6w+MWaqxBdu0SRt+iwCZRJYsxpIpRCJzygk2ARUAXwYOiSfxHohIsCLpFnpMAACAASURBVARaGLClBFMBkxaQCNw0bgIU7MXNC2bgvIHcNVvYN5w/pHEjy+9QHeEPhiDFhGoUAlbLoUGKWR4YNPOA+DWjAssKnCAIBS6RlEKg1YmGrAMccmoJQYrNbd5Yp8M0IBUYmkS1h/dK/AKk6jWuanAz6pIhpBDjF5+FoZi/jJRaSyNnw2FAdcsReosBVasiuIwfUtt0DQ5AKqqvugvRkHLQgorpzdiJumw+h+pFSikR7VnVq7hgZWhRXlceJhpi4SAYjGaLUPyqFpyAX4oWb5gAlcbMsGL3AFZ4VExtA0ip7Jd3HBoFcHKCpatnNGgYDvWB0y8wikFIhTU5/g+USqPOjJXpwKnJVLqoz5Rtg0qB4JgP6SCFl5UvKoB94c+sQSr2k1FFchw+dh6YgVT/dePBIrCCIGmQ4rJYwI+YOQL6RLABOkhF1Zb4QriUSpTfEFJmpBQg9ReQUtMPVy6o3EOLViyr3gFPFlXwuwb0fUodUni/r36CKLRl2FgHJMjB7CoAblyY4gf3DPwPu7NzD2H3hrHGBNZKywazNh81d6D41dY1Ti0d2B+7+iogO8J6oO9N8DvDREX4QAFskmhbiuOu96yYVVCMEzy5LEYVUvBpnlUWdvppDn78TGG/lGP48rwdRaU4mKPhbL8pjUdHcj1xVtRCkAL7kRQ+/S84o82ePSF6gLZbqM8MFulRUkrrZ7AZArXsJYLc1dwTogQVmJoLMRWgiLClAlJKRyWzX28tKbV1BK1cUbp+qWPo2dFYrD5TGDFpNUjNS0FKhUMK+vfhypIVHDKTBTEVBqnROMRNKr99Re0goWrFGAjgcDrz1gBrDGXjFaa1Z/ssy52H07QQDxS6E+ubiLKQoD/NqcYe/SgpBQOiV+X+nOnQIGM4GZhRr21AzJHWmlQgZakpLIGXQFVp4ZElqArGWnTXdB0IQQoIxna0Asxz7XZyUoqfwiIPGh8FKZAFdhxWF1GuKBsI89MbDEU40f0NgseGxZT/VxogdcucNrsa7gla35AH0SAX9aXZ23bjpWphpxP5IbU3qOOweai5J/z0NP6wEEMqdA8eD+yB7leECX8wG16IMCmFRQlaVbIHJ60FGV/oyeA3YGoZ3GzRkxyK3/9r70qf0tiedscNLLTcFTfUuKHiVoK4g0FLzdUyKnGNazRqctW8PytGg1WhSiuf/HL/jfyV79NnWGZgBgaDS5KZmyvDMHOWnvOcPt2nl1GXHSp+D1k9KpCLlBA5YSSsVlbm2JlNWWMghfXByPDiCPZgo4qXyJPhE0zkVFkjxYC4F5fCGM3vxD4PVpa5cA4HzwwjJFxFzKfMeoL5GVlOwkrF1CBlCjS0traQA5FuOQRQCJaiMmZGgDgF3/TkKX5gPh1d+OFevAvsTUnplbn7wqs3psH6v/5RXKofQzSjhuChXIMY9l27H6ex1JEhAdRi9cREypCCglgBKd5fYWNSOzA1ekgsS4XmOBO2OmGMyboJad9B/VVgZkaoPz+elzdP3Csp0fkUcGMBP/mBtVP92nhGDpR+rlj1BCKj5/RTFfh1fFWRknkldIq9KWmZfB8uBeEQQAJfxnQWZGRh1CY6lJCCCcVQWPJKCVJAcs/iorNxYxge57CeUEBKdKfle4ZzdTz8ekJNghpOaPzCLYSH7LuC8MNpgNSfw6XWyNrXzoP1vb8WeVlaavxNeLthuuEzbZAi25smd105qxCOYEIRMkjCcGJIwfQCaiQWuOVvWNYMPoUWm2Ywt8ZhBpDKgEqfFfxQJspbH1NC5CuWM7wts+8lE+zlnGzIw7s94ncxtAVKQxvxkafkJ2y9Rd7vkls2qCS5IMrvSHqOPTh0HwfajJ4n0h1zWQpIAYg0Hbb0SwlSgDFMykY2KWsdZnnFYVSI1qI1gbXjXscZYpNil0x+xHAp1hb55sOyZBog5f1jNH67HHsKjOMkA2FG35f3ZX9dUc7zvwap6GvBoNsZKhkdFxr7jAPmUkGOeyQ2nCqBBR7J/E/+HpXnGHpCmopDnRj/PMeLEpQPqX9jPQCNDSH7CKQpK5w74NvBz4t5H6VAGSjMZrVbg5t4byrks3gvSEV7y21PesRAKhctF3gU014KAZzBoMc3EBFhAD4g1QJSmEwkmkJCsjg/dzMBFPMq2oYmyrkUK4tq2YRCPJkGSP05XGqg52751RmdfMisOinARumhn/mF7PXyDNgoFn6R165LlgrQi01IaJGSAB9PbUEudowRjuOj39X1Hd4RCJvh+0jU6oWeNnJnohMXLWJDJfYOwaX0FRB61PU/cLW5s0o23Zl+XclKE5h+t8MYFNDme5Q2fqGnYj6AKTah4Kv3glRMccm+KiGF1+QCgxUvKzUuFTDZPt+U2Eys9IMvH14Rb/CHet2ZZ8ckF8OhRA+VkEKXISfAhIKlUnT/l5XofwyXGjz6Pt91/Nlz5pvv2uPlz9whqCQfncyl7gkpWE/IIMXCVNgEg0pfr3IOko09/1WFKwVISZoqefv4dUtcKtmQlP8ORT2LdvXOXNjLlThhSrBRXFI6gYDZoRbrghQMOno9glpPACmoHM9KYCCM+lODFG8+2veK4XffG0QkZIYUBwVn8jBVhMpcTivpHNXIuRTuRYCd8wZgGr+nAVJ/DpdaY5IyV+cDjq1sqqqQRtQhFY0pJB5U+RPLpbBO4skQLw1hTWm5dO89bfe8Q5gEDA39XAqthQlFrDB1Dy5lpqK1AvvIIlQQJst/269H/cUDWM1DupH6ogdS7Bg8yntTPKbuIUupUC3RJSWXwirYxH5TPKRTgxSYK+LRWMDsHRPvfdI2gdXKIwDyE+RVXoTHtyMOUgi7R10wSOfRkgZIef8YWWoNNGEVgTTYseoRMo2Movy6JI1fhM5Y+LF/HQs/mgfem3LhB7pjBsQTXAG2lL/l7JyvQ3nPWoIjr96FH9DU3wsXbWXVqXMpQLNp0zK92Izu08eXt4sLcI1kE4lwwXogxWFwWJpipcijQwq0ZBvFIKtj9KcZwPuC2tU3mXlzCBei7en9K56fGmvKd3Edr4gtTtXfKi4ruRTELaH0C0HqVzJ3oA1w7gAw9R3P3HoCXCpxT5hLtSH+uew+RIbkeCJAoiYJWJOACLWyhZ/iVsRK+nQz2IQVB4owgZ66IYVxxBsqPJCitd+DS6GAYFsj+wpC0H67jdZxkdFDD6SwYpL8plhGf2wuhaZiW2HohFWOKUEKoTEzfTkIIwH7j+ML5Iz7uOPvqbLJOx8lQ/QsDlKgGBKSSSYUaVBPeH8RUs/Jxi8BMJigEpdSQIp3zrFtoTWfiacAqVguFX09YAjBDlgeIGIcyk8JUlhr9WPfP7w+k8pMnUuBGzHcWRIRJ3Hqcn2QwqBa89uFwuwpICX8pqBKSAlSobkxGAQp5z4cTzpLqiHOak1+kZemAils99ZCmhIrz1/e6n2uXIq9riJHgNpKEnuWIhb6fbgUdD3zyMMnFM4RkitPknApVmGz0CYYHSDl1c2l2DwNe1PKWN734FKwu2FfNwwHzAzWUK5eWR90QgpuFoj0gA21J+FSmA3Lhd+U8OpNxmfCvUO6UniusRMGzC6/7lQhBAAHOAj/rPGpBiksGXJyWJb8Y7nUJa+Lowe2yS1pgRRkKXCp6LoIlKzAlkQQlNQ6WFwCl8J7UD1YnGK5RXDIlCDF5qqSQ62s4HtxKVTPBxfDPi2y4vhUH6S4NUfY3sHjKXApVApA4w8YpNAF4FODUPJWKdUT4hfkaHTCclcnlwLFuafiX8CEIIUI04BoOdDJhAkhry7mXAVSIBzClMBBHQ2IV6KLXnIH+RA1KJfW8vKfpSwFJryy09i9ND6+tCT9W1qq7ffDsjHBy0qJS0ULQoxfms+JGHjJaRM+5zh+CSAVvk18pgQpHofkELtB0Y6lzqUUDVD7ogdS3AAMnEo4aKXCpdADVqpiZLPmmuMgQxMapa5aa8Q1FUiBFsh0gWJ0WaKHZjAoXgFDLDLERBJV76ErCQ5VSIUt/QKqkOK1iOgfikXEiiBPIuo1PEsbPyzGXk2VOZ3ZQ9NDZc6h7JIShF6ZWsyPWSIpe6QTUrFcCqNBsCkMJK0jCZdSPAZbOa9c2aD4Me4LG2EPsJ2v7JfUuZTsYfVTvZACKYTSLwUuxazC2thgxXh0VSD4KfDEaY/V2yG7qgapIL1yBrH00gcpBHrqrIVtsslUwBb9gLRbZVNXVqXsVAtSwm9KDVL8rKkCEXlMVntRPtzwqKVBK3fU8+RSJpe9yG63W4u67W58dI7b3EX2YGfYRVZGnOgpINUq0+RFf5CdRTR+0XkUY8JMxZCm5MNa9ghOdXIpMY5S41KY6QJsnhb2VOJ6n4hLiS4DGF7em7JbOhQwFz+q/sFaj/anvZZtKuqxONd5Yy7Hq0PfpQYp9puCKb0+SGFHsMsyOtpMbd4M5wSEYZrZ0dlmwY6VSnRcworZSjkwSFeDFHpZaxkt3Ski35S3BAEBqocyVlQJwi/w+Xn1mpCrZfg9du+WRkcz4Dl7Np0xArGzcHo6M4HEoxNS4FIHciU6T7NUMLwESmgdIS6lY+7lBfmpV7d6giuECh5+U7Kyn4hLSZ1nD1kPgughfpP2FCOnE8SLQHMjsuTRRLa7+dsGUuW9s2BbTH6P2rnMXyryM+ydZ0ugsdHDpbDEC8B3uz2LLr20dl0L88Zbi842q0IKF4OIQoEMPUGVhR86VJCXbx9dpdEDc2c+LfubAp0aborPkUshUMVWZSGmq9Jiqhulzh+nwbJqmr1p6bje0handEIK+1IH7fIhDHnTTPMj7AaiMRCYS72H2yIfQhGhdSMGByC1qH/hhweYMcBTCa0IDa2n5FLcGvhNmYr0cynexrIGaxZpYZQas5eored9BgefCHcnAhjliQqkWN9AFqSG0AUpNrX3uPDexkZoDsH2lo4P4RigrETzm9rCD4twTK7tMKGIhxS/HLatXcmhys1ZmCfP77cuYJZXPZ4jl+IXu7ePXCTZtdTdaz+zwKy8lK3vaXMfRnuq/YDoSPt6Fn4wm1VwKS4NbnrMpjDzqRbOXAqQYj015xbkQ6MNYUhp/qzyA6Qpvx1SebjMJ+VS0H+yNNWin0vxOMycutigQE/2+XeklziZK0GWG5WOKi6pQAqjlh2T3XoghVeCei9v3lWQLWPo5gQua0fLQ78EKSzCsbxrht+UOqRMbqSY7EYa0LqLPBq5/uz1MsxUjufIpTjb4mYh0Rai5DX6a7/sEGInUB/eVw5yEal0QlwSXCoqI6neJslSSi7FNwbpCiVH9UUxD0uQ4siE/7Jqi5VLqsTEY5J6Iub5hF8hymVV88ZY6K6n5VLoGpR++bohxeMwt6n28wh8kQdmS5u2yhoXMBOGO6PZc1VI8aiefouQjwmFZi4TDA0rhsaGkVWEOds4HLW+mvaM+e3haUmz2tAPalxKFAu/qTpAajlGDMD9EBLHexfIig39iV7qG6WWXoTgV6voWXIpiFE7q4AUZu9Gf8FBHxbofupB0ioOtavWC76mE1KxshQ/irm5c2opxmqdfwgdEqSYfuzLhv8wnWkc91j45Qpnn+fBpdinEhEGPbohBeVOECOr4caD0Y1g4rvOvpJPvvtCCiN51gku1aQ+WKM0F5DC3a1lrowJyj3vfz3kKPtUHb0h8ZkWpLgn2JtSg5SZlkqr2C4D4uYN+ZAqp7Q+BnihOp8nl4LBFvQpjdkd2Am1HpXlUmYWXe0RZUEdpTUB6oRUnCwFQiCuHX1uxwykgRQJUutli9Nn1F9WWX4J0UfjnUmQ0vhR9TJrHNm9N1zik3IpTP/sNxUs162eMJPne9VWViVNnPcv/JijouBybPR7tW6rcylU75p+VTSVnEvhlTROHm2Xj9Bl2VbmbQEVuSb9mm8ltgWakMKM3dVO5bFms7zZ0fCzb+3FRsuX1l3nPJ1eV02exyRMC1fyHLkU0jVUDS2e5lNX5VKGjzxlg/03ebR9s/z2HO4T4ZbHfuqElBqXYpvJTlj6BTW0fnh/SKDTV9z4ppMW3hXMLcmV3sp2mKgK6onUDgwvhJ4Nr/yeVJYCA8aIyO5ESi5NQis6x/LninfR10jWsUpvHu9FNExqkFH+oAaksNY4s3B+qWS1I0C0HeJMYRG5VysX1+C7RhsHyZljqAmakIIU33zeyf2QN1ZQZW7Ed+x4Ycvxjg4id85JaVa/RnXPkUshMP1mTtdeJ3mKLZcepIzts8BhlfIyFpE5VHO8AlKc+zThwbIU1BPtMcTAQk7sTbF5m9oBQgNSPSO7bUSz56cYbur34VkJUpo/q5XOD2WNscZR/Pq0XApNcJF3pW8u6aCWugJdNgRL/JPkS2zzYhmRDA94VGVfSuo9IuxWZyCyqVS85l+8Ml6A894yGgwzJDbSTPZQpDTcK4KOxb8nLHzbr3Y4XnrkZpwgzFboOyoBy2KLDdQqvyVyznOSV8fOXOgBVeeO9Fuio7k4oF5Dq/GXGy/sXfhEvR+4XYKU5s+iB2yJrgIp0Mck2JSSkOIR/gNIHe7Retbmj1fIluybao+qvCP3hE7uBSnkIIqaUDwtl0I3kBH6fJoTR+o5kIAiF2lCrUg0YQ3kWnODfKLjUdjgS/mllASEroPyrsua2AxZ++BsGMhrYQ2a3Vx3MOgWqVSR4kL7GcUvcJ7XgBREgIay2ypi1+DogWAeqBFVonNIAmM1IzUqn6kd8Ek9fYaW6CF1dTAXLhP/B99MBExFTh52q1O+Afm3X4EUMCWkqaA6YplLHRNyWL7vtVlt1Hm9pskO7wUplI/QCSEW+eRcCgsCx1ekypETN+3n2M4vVaE2XnEgUP5NJAZNe52yAl20N8thU2WXpNNAEAPhpQgyFfebzgsm2vZqSwaxhTwWl4KC9B94PkC/JsAFRsVedmwHHNui6HfsSx3xUj7REeFScSs3FneHkehR1YyaZSkk0IFRyq0Q5kZh3alRzb0gBQ7bD5EcW17o35NzKbRm99P3+rOzPH3HmXTwzWd5ep86q6rJUIEU5jYz1X86WUZJyWoXt+Au/kQTRPXJngn9fpa37M9ThRQvYDt/Xp2q94Mr4goTHmdVl6MJVlMxA0cVUg8QE52hw/hhBEVgJJ1ogkps9SZiY1yWWPh9Qay/eOhZqRhJcsQaPabXPMzfHrsW3jT0OWjttHbih7abiYCUZhvjCpYusE5gZh1KP27V03MpxIqdKB3N0H2MZvB/oSNyEr6g/jlqmV+K7sUp6WIaK0XOytQOndVKhaK5pTudWhKxifJKo/2JbUbyimBF9zE6apVdi/+mCqn0y1LxFeu4IhZ+vwApNkjvZkypAEJwqb0hywhyZzstGVj3qdwk2ihBSutXrW5AO8LuvcIq4+m5FIdIf4wDvVY94qc71dt+8SKIrfGa4pYwqdekNojUS1GF1ANwKfXaE18NwvBK38JPlUvxOLo8BqeQ5zoM1yiU6AGIUTDqDHpYy6XxNiDbQ4meLNRquNjIZwDlwoRC8M6n51JYZctiwUQamcYTzFuuYC7UT6oH+0BpElj1iZQuYiEimcCIRYHKo9i9jqavVPk9colXiTiU9mm4yFpPvYcqpJ4Jl4Lu0g+X1ISdcVEnNH4MqbjbWI/Oe8sQaOKpIbgU5lSQEA/CrTP+ltAVJK9e1PxR8wf2mN9mh1rc8fRcitfcIju9Znvv8UPMvP0vFrsxlyKFcvXaFI7cdv+TUAXadeioHy3kkSKpzHh3DgYdolxMt2pDSKu1qpB6JlwKfeTkEyqAiPYmAimVu5hNrYJNqa1GkDaTneu130GkCkBqJvIllRPE9EMUCtTw9FwqlWb/1fd6OioQ6SLYvASGi1kRRqBBZvDsLqT7UIXUM+FSEPEjUb21OpQAUphhEDJmeE4M69jnAaljFc4Wexu+34tLMVQ5QrpgU8+AS6n0y7gURwEYgPw3lJtLHZ/eeSCDw/ia5QFEl2Gtpe5DFVLPhEuxa8Did2iAExwJIAU2LklTanr4EJdKUHL4p3tzKVb6cZIkg0uFSfncP/Gq7+4QBH/l2zsPsl4WwOTDZfNgV89jT6HpqpB6JlyKB+Oun73ftI8QpLpwa9wBbg1pivem4ouA9YQ8qlLcs9ELEqQStSF6r+IMkBJRKPD5sTS+BYp7U/2iJ/ZEqmUa9yMK/dTxCBVNF7+z02nZ0Pmxh5qRMK9iWgokro9CqpB6LlyK7YaQCiPI1l8aByAFG78vqpBiodwslH7gVzElwEdGN6TqIUvFPK7RGsVltt9wrENlGeT8UvcoQFGa8osBKSU90vMNyt3bU79tdnH2h43mOkxFQ/CeWHnnydjUJyNIrVCF1LPhUgistMY5kmIBEaVgZOGnwqX4rgA19c4xm4oZ0o8AKZ4QOHsvQ8rgUtFX9nzPeOHXOHPSNVl1Z4Nz38Fk6QzWOXsXpSJNV8wI0uyGKqSeDZeCtoWlKRguaXVA4lLf1bkUa0RzaZU94mMUgliMgUtpFyurDquB+3IprDw5M7oeLgW7EqiX2P4bRtHorQt6pgS80eBSsjeUtlNwqW9Nr3qH215d2Gj9OudLxiK5afblF1UNl1a1Jk6LpZjhka5cL5dC3HB5uSaqq1GWJf/1PuesBmeln/ZGm8SlNCHFnKINbCrWwxCXU4PUvVoPMPVD6Zerh0tJlAyBXNqyVBBX2QADUkp6pOcbvE0vKtw/junkhzVwfUr0uZKQ/qPuei6xjkxZO7jUPLvfY7c/dAQRq1wEZ1feGPsNAVPg/Rzkh/Asz60YQDk1qDvBSIgtJOl33mNDmvaIg2zcA8m4FC+79jfj+FzKXCquYj0X2H8FiTyQVjGZLAUM2fIGv7SRtWryYKIJgW/GBpcSzCQGpPTQP9V7sPD7iTCCTfTim43KjltmP3jJnp2DIDP5ukc1O1GvFyo5i4s8Fn2QGopRxiGx05gwqUsnqBAYgFNhaFEnKZcCeBr9MKFQtilVSCmf1mpMzHVezMFvyo43lUyWwo2zoysj/k5770xdTgF1++uuzngi0KjYgFQMrdPy1URbPYg/iyG3aaON6eyuwRVaxmmgfRbbk/oOjtyxPlVduL+6Wri/v18o/q4UD+uCVMVt8coKHsKD0t/C6neZPAy047Xra5XiLszfiIoXVFyTfUnCpSCOQO8m2JRCMnmchR+LRGzpl5xLoZkeRJHLOnBjhYDGjcDZCq7imnZSBqRkYyB9p7CSgNcR/oL0ZG2C+afJLcwHrZpzemzliHZN4/M5vrrV/fl2X85V4aqvy5dTd2LFNqnG/BguAsysBo/k+LpyunJycrq6cOKr26DcIJqkF9LhwhJ9woTCb9fku8m4FBQUiHY9vBFaDIc79SiQ4l7Be224hSGVjCQBqCfszjVbr7O8JmCdKV6sjGOtMioZkJIRI32nMJDlJDwY/DBHQuikXMYY0pkK9VZ47CSuDgq1FuSAJVNrHhJQUsXJLns0ec4O5uK0zvKCtJYj0j2AuL7a5UVqn2OiRxw6NX9NfiYplwI1cqmwT9jzRZuFQicVEWq16zfRcpb2r0l/gQnFIC0nk6XQEUTu3mknT2t//c0h9ZZszJdhW02reANSWpT5pevYAGWGIBZafIqRjC9sP8scRvNtyOsEm3N86qZcR8axv5PWbued7WbEC9zMKUyoYAbu2IC7aLXfFbRmer8j+8zGnuMUcnhT4c5qUSrqEXlrNM7hkgo2pbE3JXGpTC0lOugA0sDSb4P+BxPjCFEeC1IcI3i7N3BamoxLcTjBvU0EMcEDE9NUdgDtD1IFaZAEP1QhYm+kO1q3GdfvSwFmLeHjX+xowGhAy3MlfJv4xE7ql57SLap/Z6OsVSqtpqLrXepb532cxIs3YBEOgi8HOVXAsnMQ4ZJXJm93kYRtfrmmDY8rqvnFL/DRgNIvkkxdWZrEpTJzEg7ZIK30CBPIKJlCXCp6QVms7JvEpaI3arMO2UOyUzM5JvuTxcvHTOjuAds0uWG6nuMgxzxCsjUlgVQCLiar3zj9VQqwExXEB571kh4B6shoKH9D1Q6isRkqWSA6f+s+r97pGU9igcFp0ZZHclZhQrdBa8NUg93Q9h6qyYIiIVfsUiatXPcNAP7uFNiUaodCXCoxpCBN9UKakuvP7gkpsQAAfeWTWOKOYEIYmH41Kpv01O4H1zx4OeOwnHaU1VX2jlP/zfEUNLGaU5PgUqk7RqpVbVxLRgHk22oym4rcOiZTrEp6dq0lS7Q6QjRRSSv+/vUPE/ZPJWtd2Uki2oBH5Zc3FxfTmxsbFZQ1TY4CUk46tniHP7MspaP2ZB2J/M7BWzmVn+r40sWlkGZoZQcFoIQwLu8JKRQQLiHSvsQn7N57XOIQTjeQddUPuOYEKrbXWlsrPKfvF7A7Rc1vj6yIicXistoBgwwLmpJiWxK31PhViAlSegimO1MXnxjrG0M2zueYC7VF+Kr6Kg4RNq9f1Fxf0hjkbyQ/sa5PrzjfW78tk30IISsTkRjVXGaSb5X6zz0wXO22D7evX5QHHHfdDcMn0CX8m+jhFH9jt6e1YZv60i7KpRIML7S2pRdZd2SM7h6QAuO3DvTnYgnQcQplDiLPYX0MqLCKKALVuL6xb/HGf1naN8Q9oedCgE6xlhQyMxYMwkfOCgsR9fesp0DjHqaANIbgnSwGP5Y10icN/LDTx9DOEiay0OSuMqW5aGP/0nc3YjsddnMWgKALcnwHlXyhWnhEqDKFMOUh8d8sjpzfrjWfF1HzVAUVrEysOmgP7A4mdSgovQfCpYJNqZUpcamaHI11ofQID+vCHUhT0RIEpHR69eaFNH5marr+gBwx9h8vN0QfQXszY4vjfGpzDLyBvouq5aQHx71aXhZ/Ih9aD+WtfbZAvER/2Epf6KYwo+Id8xHtpXGWEgVc1F2/4StshPRSUdg1C/051fuuuom2b+20VktbrVu+fZgM0bavcOkVRN14jDD58y3jbF88MwAAFQBJREFUZN5Z7IKLxPbxajaSaXy82Ues90T7UniJLtvawsfyygrX9AFVTxPiogRL3tP7YXegBIM/3W/VBGkqxqAwRKsIl5JxoHgyou9NYFMyZ0YJUnqitESU6IBPo//uiuhk+BsiLeZv79rJ5C4IBlz2zgQDGXijip4ZR1YaD0eWw1vPwiFvfM2tIJAsda634hXDPAwXxQo3IUHiSWRcYbeF1/8trleet1HDtaP6Fkl0qu/2u+76aeDaQ70HlPl1b3AIubaOvvn2s++6VVkH1gzuLTtcFw+rO4nyD1f6KRikuf1Z7NcmpDGvdQK0Oka0W+Z1bpE9a6asEBZ/7UPTfR5X2rkU6oLflIzJRBqnj0tBCyqUflGo34NLwQyjcaq4vIVmPp8P0FJ2acn5G8odvSSPU3vfTAhfUL9Ga440/ddOeFnPqAlQzcsZQOryJQzuuUgxF+pXn/xaK/6sp5Ff7j/syJat0mYGWNPPJc+3M6SOswhIOQ/p+10jFdxtUDlyDb741BBr5cbE4C0s9ieABI00RELvHgyK3ZtkOngIEUFrID8/ACPbXbDCQEH/EkpzBec2kIo8EYe730swwyAdPAFcNWZdw1yqk3jhl6Bg3hSn/KltrA7Dy2AzpybQNZFHuBSqaPqR512u8L+526aWAhetlCB/wrfTLic22BNVr7ZCSHC/rp8g2YnOwEjzW0kHUrLBgNrUVPW6CnZN3Utksg7AUyvtQNbVtN/2pgC9mII90aqFhtbJ6h5emPtRgPxY59Y34FLZh1RdSab8qSO6hcnfEnuLq9BXin8GLLBRE1YNABRwIRx3EtOFN1B5cONmjHJpGxZ/MXKJt8USP3yPX4F8sCnI4bHPMpcSkIr7JeZO7E05wEZDO8ZYxOmHVFiWwurxomri+Ptxy883RP3F845bm4nOPk11qhI3pgEP9BW2yqVQE73I2ssiunJm3ThsVIWZtHDKniiU2gO15vcuNkAnvUErXWUQ3Hnd7vPZjgtA6vW5aevaTuBS1dBr5/uP6OY10dxtIk+Be9OBsYO5khEU3hpOsmC8b1VYdLX6bYKBKvEa4VLJIGWCpd82q0HF8ylBKmyQBEj9eNXm/3bU9HWbFq5XZldukdhn42UppJr7duyXn8Pyu3zNSVkv2mfIWoTU8bfLUBGV5QFVwQeY2365vc+5gACdfdgl9/k65ThNdPSzwX03AVvmStoGpMomaB3uU0VTVdRXaaL9T80POpE++LtzYYUJE4pg3NDVyaUgsmNvqg98TlqhpQQpGZe6fU2bQ1QL9UTlZ6IzhCiwD69iStOzsf4wgylA+35r6Yuhth7sti/l9PXcsW505uVreME9+Gt5mC49WalY5P30FjvLiqgt2/L5eh2aqAvfzHk3ZCk7ZQNSXnCpqVPEBS/LysDC7+neehpIxHFZ2IRCprOTStXNpSDt5PdCnRDiUrBWgF2gkuOptxMx6rGk4gM7B3cn5MGuwacB2r/Je33zzUNIQ77xMy+hKCc9/UB/AaleWoduIisLHRzZ7RhCoDPT4ssqKJr0dO+BmvVbFhug95bOwQNY6pCdVXWQY+ZWvjSRK//ISv0VVLGBmXmrBfH9+hvW2F4skQT9G1BA+E3FqedDXMqH8Z74YJ/79T6heMZIA8A4qpKeMQdpCfZaOLDCta++YQ11U10teepK9s4Kg7VIwEmv4d0UWlEmbsQD/BqglSkaHzolB3KQDNvJdo19kLqS13fIqmpwqdQIjqjew3gCbzgkzEiKWn61uMDU5M0SJPkd217rHQGTSjboUqv+0e820VqviFSjqFniUmN6ICWxKWntmCqkBPZYac0WCiBv2IgXV1hfyh41euCpaHp6vgTochhv3kzeUSp6t3jieLlOb3+OU90w8oY8UZvS07PHLwXu1yOsAwNw4NchPrFTwQDDTj7ePjb0/3UFseN/PFRW6E46iz9+B1KtEXtTkKZiZl6JS43VJe0fyGSmwawQZ5IgpWvICS4lQeqfAHL/il0H1o9ytI0ArsAhKwigpdqdNN3vou1ZwBq5L6tc1JHTs3wE1ckumTwTDys+p6n5z6sY1n1LAyzElCRbAEm7HfpBmPnhXNfgeV7di2tNUPhNxfQEueqhRAekki1rmVpCmpKEdiFLxZQVV6O4EFn4IUt9WF/Ipn28AcEH+BPzKF1lqdfwa1eBdoAbc2lkaY+NZTSQdx1/reS/72l+pf9LzNpdpn9AbjOS4P0B5AEkFqshDCkGSghSvqSQwgoZbGp9h9U04FhCltIFgyiXwmPhmUqciv0DsRJEo55MVMUahadWnjyhFwW4zGJr0BSUOvoHvPjH6wLPu4rhpVI1SI1JlFf/v/8B/cIugreiN7K+SJCqSS5L8TMmys8ewNIP9DDRd/hY6SELNH6ORPZEIDDWgvI2yZr3aKdihcJA5xrxB3yKF///ky48WjP+hoqYuqKf0t/fucuYh4UJhQIIES6luKrRTZeVqheZz7HaRj+koPET1OMRi4OZHK+uGZn8JWCyLsHDpQCn/E/8jDONNjzSZUAqCKZlGFA8Er1/z2rCoWcV4JFDKuk4hg1Gfi8sgwEFASn9SnQGCIrHagoLK5zAgCvIzF9aVNvOd6kO8TdxiS3B2bwLn09KZcYSVJE8exiHQQENCoAtQFWMSIHyURIDKXCyEE6YZeC/8JASnxwhHUo/eB3/mwqkXoW5VKiwUKm88OamBsiGsBvj46iNDyy0xae8leLCY/xhLskHFqLNc+TZLRLs9EmaIjXE+PucKYCBgTR1vTBIl7UyAikI5/Co559YXGdNHPRwjAHWyvFIA/dgDBTBIB18JhVIhbZ6UW/TxFqd4wybQE2XfcUVcKmo2ut5C0jd7NIa7FQO+307pzAHmhvZfLWAnXZZOx/ulKUm/mdC2DJwUbBI/A8dBTxCGxBxgyOBcOQCpolqitWHa5lR8u9AAbApjhEdbaoEqWpo/HgxRu7Ws13OSoCtIowhxNJ91QrZCTLFAjLicFw2M43BID21hd9siEuZqeGT/2D96xkVZVvOZnqb6Ozb4OHFKnlu12gzi7b/8769+tlBDRd7rx0fToC4xzgk1sk6KK4NiGLqIAQOIOXu5FTBfJWV6uJERjpxwfjzt1MAln5gU7JxIedSVlry30xPVbbQ4Hfe0mV9O3KJ5GKQmUsm4dmB58xkn9oF5lLhUuGFn5kKfr4g2i+jyesgmYerKbuQ4EnRlAtI1fXQ1tddoukxmi8naoQ1YJxB4kO8Pdil1axmV8I12zRW5qy2Bqk5y5kxgGhwDmorLKAvE4V+Rye57F1T7e8/PxLnfIiOGmU+DAWgSGc2FS1cxqX+h/jjiGSTv2sn77GnyVQUsDdbPfDgp7ZmqpzAgqxoqQjBh6pnmKWxxk+nekLGpa7HTdR6R6sZZKPNzdybPHLV/tgmRD/09SC0TRMixu7TzAicbcrfypsZbXC6zwJU/GH9zQ5y7uz/+Fh1/Zny3+1tFf7sptUsarh9Q1lfD/uHYKToyN6dvRg2IJVu+v/25WG3li39hHAkOhOCFKwnsPQhPxsmIf/Cz+uSSxqZX7wuWDkgOniXXVd2SHQ41du7APP84QFWhsVACtIGazT4n3KTyUSzfRJfxMLvRwfR8gXtWwDNrC7TDYobv+sIgEv5dpDyEEbKjkLq2SGXrff940DKRL4SIvfFx+AtGpN3R+9voR6x+Ogyi2ohS83sIZjBjafgDoxrpdeA1G8PgbR3AOwlS25CEYEU2wzQlw993zvAyfq6rC009GOrk7I+U+unelPmyxOqv+inj/DF5CiggFQm5C+ZAgHaDOjAWcxH9DzZyhKDMAypAFV8QjSXUQedfpujZrh0zMDNc/8mYLsWkOp/x5Bapdc/5+j9y9nHkaVMlAOmmJs92QinQ5r76RkrBwctrqRLBwJdbXAUYdq6sQ386AzSAriUvGtpfztGgb8fBXhvandKFlUhCinWfAWXs3o/7dnoeB7jGXGlc6lvlXwIzZF7PkFZoxtHGzdwF7SX7YLJfFdAihVlZLYGXEFYyEla8BB1opACl7qt9Jb1NsNt9iLrej5Atf6y0utW8twAUj0iMS/1rFCg7sK56X/xaFxqEeqHm9f2uzUOQRL8MgV1Zk8fqycYUo59QOqdZ+kO/PXLO2bDxmFQQE4BAIe81bxbK12VQ8pkthJZ6/87oT0fcqwiHqGbIdV3TO7g9CSVnu8sZpXPAk0INx0LKSz23LVuDo1nq7Uphp0CUu8aBuphExUwbcxuQ5FI9o+vGpEzpdtOjQVkb0Zsi4omcMvGgjbo1RXIlHcinecmKv6vn95+q6WZDLd7tJK6f57QwM96ulqk2vMt8l4Svblto3Jv49aN34BUOmn/+5YVmVlZ0sF27xGUfuG1GSCFCEnVQpbC7gt3cgoxo67ApconJUgVT0PWuD5krQEGOXanyN7bSoJLRbFjog5LeQb4z1H59LFbrlOMQApyXPMdEqqhbuAaykOTUOcjpYPksoaC2WbKRba6gzOnhd1AHuEApIayLDezUDKWO53lBQiHdTN6vmKisRyqGO2mnEyijtImarD09vnKDEg9wit59lVg/PI2Em+8YDuT/ZTkaRGjkIKYYJ3P7O6+hBf74PnaHPkzAamsKyyGJlouX76g3U+ZFUtfkMk4l8a8JGSpCKQw/L055KtE2oSP1OyWEyUCKQCyabUlZLwBRsU3oXH8CZ7JcOZTaDlyD3v6VvIF3viehz0C5LsKIO4bUG7eeoMdObjzrxWgco/HlVsUDHgQHd9cBHcErEOLdwz1xMO+jd+jdGjgzFhQsUUA7AKwb4tEHlETigik2L/C9SUje6p0GfHsijMGafUjZKlBbCS9ys4u/Iz84Kel2UOb2MDhvam1GEiRfXqb1s5N73s23iOubQRrjJOF0Ehk+IR1jZEb5LdKBHUF+RPtjdwjXX+YvyZqb0d7mUCoD4AWHJSJwUAXJiNotRmdWN+BPPUoi9GH6alRatooIA1aYRSAgQoTCJfwm5LKj8pSGMMBhFVuwXTMM7ZgahB3YKRDthYMNJhSBFvsDApcqfESHEKigx5cyjJGNT/t3+9GvCWN8skcXGpHWgjyk6HFpXbnMK5h+MOTwD+PAikXVSFyC7rMFhLCRxJmJEFGFq9P8Y+ZOzh755hv1fDz1X5vf88vPPNaW8fWB8jUfLA/62YP1rzaAcQtkQYsQ6pCyFLk+gcAwoAOYnWIsYTBBLGHx1bQBSNxYAmpO8GiMJEHyNM7d8hbveFRj0un/ozK7PzvF0WEqO6yzeQopNiPI/xAojeAMYwbH8uvginEe2roqbBpxFfuPS+WxVXBufArWvw40l0i0hi/PT0FeN+1O+Oy+uYVHe9MTB9jLXf08pIqx6D042HCC78KWp9nePCBUST+5xEmLvDg5gEGaYcHF4OIo1Bkbr5lLsUDTRwYgE3jAyX0otRFfftKSGHhJ90nng4/of0pquCmPc4B6QldkqQ5UWPknIkXOphr6ZsQwk8Yn38sBZBy02ri/IpYtDVhp7Zxrw/btzChgH01j1vW+AlIhSnAszVWh9IAYm+PWNbCI62ofHGVApHMHUDU+HaBZZ8qfrzdvq6SixxClnoseIT7kNLns25cSj0xbn4UCnBsBSvtIS+IlepLPJQzO+gjmgmFnpUWfkpIQZGBOZuR5IJjFKSa2DEHM1qYN6ySFb9JB7SAb7yWQauVTr1QvsPRMHIE6XWPJFvFFvMo3TcqMSiQdgrwDu5BuQ27s7WIR320SauwCupHFIogxnhk4SepL7hysCiTua2IdV4soLN6Oe4wu9ywMFX5Ie5O3FTfJ1dXxN9hXDEo8BtRAHIBtOYHJZy/dqnsFSzsSied/g2kRYQJhQQpqCcispSQGJClrgS3Mqsy08kx9mziOgy+d+V/dSI/XpycvMf3Fyev+UN2nI0kTrUTV7hxwaDA86ZAgMYsYEm05f8IcJwevC0fanbBIB3RVGVcCvotseCD43ou7ey5bJCmcDZ3/WFJhRshB1Ztz4zevIRv5OqK500so3UGBZJSIEAbXy3tlTU0/a19pqcCo3tllbDJxEo/ASlwqcFiaO9gQ04NVyP12Nu9837BjfDqtVZWQ6MRv8ATW6FJqw7fALksfGp8GhT47SngchXN9be2dpiX3hyt7XpcVlNnI7Kow4SCrRwk9YQEKTCyi+PBi2pa+GGphk0OtnT3u5q+IiVgHBEgY2FfVNfBeoqoMjquJOOCQYHfjgIhhZ2EC2ABu0m8gcsmFFjtiX0pASlOQdXHjusFlAHPX4QFo63e/Iav42qQgpZd58F1AbnGYVDgz6EA50kHpxBqbWjGwVswxM1Q+kGaClCjM7zwI+v5AgXZb6h8BboLYKH0EJ6CzWqQYv6m93BRZP9K7yPGfQYFfjsKcEw/JCeDbxKsJ1ZYlkLU12HEUCm626XSQchZCGX3018y9d85AkT8dt0zGmxQ4LEpgLBabJBuRg5ICVJsor4Jx/WFi0ayrEOScrls/fWnh58mGw1IPfbbMer7DSkA9QL7TVmF2azEpcw0fu7tunhLgkuxxAVOVQt/13iN32/YYaPJBgUelgLYAqb+qSLJekJACjZI1PJl5Q1+2Kol5KuDMXbQZT/1GEzqYV+FUfofQwETK/2oDftSEqR4lwpHrJXsH9NfoyMGBR6YAsjeO9wiFn5Cic61wf2OESU5cfB3GFMInD1wU4ziDQr8ERQw0+Igdcq4FHrFyvEoiMRZ9Osf0WujEwYFHowCAVj6uZtCGr8Hq8Uo2KDAX0MBBE5ZzPT4C2jlytDq/TVv3ejoA1IAkNr1NzsNSD0gjY2i/yoKcLxXx+eMBqHx+6t6bnTWoMDDUABWSdt3ZW1UDYOkh6nBKNWgwF9GARP1fTjrcFw9UmzXv4y6Rnf/Qgoglvmiw1v5hnLjnaL+QnIYXTYo8MsUMJlrt/I5HpIBqV+mpVGAQQHYSVBNaZ+zngNlGuQwKGBQIB0U6Mynk3OTYXeUDloaZRgUYDZlb/lSaQRcMcaCQYE0USCAFGVft+BxKLPsS1PRRjEGBf5GCrgC+fnfZ4KPlSfjbySx0ee/iwLsudvZC89dw97873rxRm8fhgIu8lSf9GdxYlwDUg9DYqPUv40C5pPNxc9tLEsZh0EBgwLpogBySBkLv3QR0yjnD6LA/wNTkphu1IBldAAAAABJRU5ErkJggg==)"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Informações sobre a coleção de imagens utilizada. É uma amostragem aleatória do conjunto"
      ],
      "metadata": {
        "id": "TqbCGYPXBXdB"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#!pip install -U keras\n",
        "#!pip install tensorflow"
      ],
      "metadata": {
        "id": "6dfyHBHHhQqD"
      },
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Leitura das imagens\n",
        "\n",
        "Fazer o carregamento de coleções de imagens que estão já carregadas dentro do Keras-tensorflow é tranquilo, vide o [Cifar-10 (recomendo ver este link)](https://keras.io/api/datasets/cifar10/). E quando temos e queremos processar nossas próprias imagens? Aí é um pouco mais difícil, mas não impossível!\n",
        "\n",
        "Para viabilizar o carregamento das imagens em uma estrutura que permita ler as imagens do disco e processá-las pelos objetos do Keras-tensorflow, vamos utilizar a classe [ImageDataGenerator](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator). Os principais motivos dessa escolha são:\n",
        "\n",
        "*   Facilidade de se carregar imagens diretamente do disco ou por DataFrames;\n",
        "*   Facilidade de utilizar operações de pré-processamento (normalização, reescalonamento etc) e que podem ser úteis em uma etapa conhecida como DataAugmentation (veremos adiante)!\n",
        "\n",
        "### Organizando os arquivos das imagens\n",
        "\n",
        "A pasta principal da coleção de imagens possui nome \"documents\". Dentro dessa pasta existem três subpastas, cada uma associada com uma label:\n",
        "\n",
        "*   documents\n",
        "    + advertisement\n",
        "    + email\n",
        "    + scientific paper\n",
        "\n",
        "Nesse notebook, teremos que dividir a coleção de imagens em sub-partes (treinamento, validação e testes) utilizando a técnica Holdout. Essa técnica foi escolhida pela sua simplicidade e para facilitar a codificação na sala de aula. Entretanto, recomenda-se o uso da técnica [Stratified K-Fold Cross Validation](lectures/cap08_2_avaliacao_desempenho.ipynb) para avaliação do desempenho da classificação, visto que essa abordagem retorna resultados menos enviesados em relação à amostragem realizada para obter a divisão treinamento-validação-teste e que evita risco de supertreinamento *overfitting*.\n",
        "\n",
        "Supondo que a pasta \"documents\" esteja no Google Drive, vamos montar o drive e copiar a pasta para o ambiente no Colab:"
      ],
      "metadata": {
        "id": "fGQCcH0fCUlt"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "id": "tb5JozJiFDBb",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b355991a-8073-4fb3-90b4-0a20ee56dfbb"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!cp -r '/content/drive/MyDrive/ocr/full_3/' 'documents'"
      ],
      "metadata": {
        "id": "SbzoQKYgFj18"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Fazendo-se os imports necessários:"
      ],
      "metadata": {
        "id": "_ZpGpiGiHeGM"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from tensorflow.keras.models import Sequential\n",
        "from keras.layers import Dense,Activation,Conv2D,MaxPooling2D,BatchNormalization,Flatten,Dropout\n",
        "from tensorflow.keras.optimizers import SGD\n",
        "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
        "from tensorflow.keras.preprocessing import image\n",
        "from tensorflow.keras import layers\n",
        "from sklearn.model_selection import train_test_split\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import os"
      ],
      "metadata": {
        "id": "y6JFUcrWhSaB"
      },
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Leitura de dados"
      ],
      "metadata": {
        "id": "mMlOC9NtJQNF"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "path_documents = 'documents'\n",
        "lista_subfolders = os.listdir(path_documents)\n",
        "\n",
        "dataset_dict = {}\n",
        "\n",
        "dataset_dict['filename'] = []\n",
        "dataset_dict['label'] = []\n",
        "\n",
        "for folder in lista_subfolders:\n",
        "    lista_imagens = os.listdir(path_documents+'/'+folder)\n",
        "\n",
        "    for img_file in lista_imagens:\n",
        "        dataset_dict['filename'].append(folder+'/'+img_file)\n",
        "        dataset_dict['label'].append(folder)"
      ],
      "metadata": {
        "id": "eITuq3OnKdEq"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "df = pd.DataFrame(dataset_dict,columns=['filename','label'])\n",
        "df.sample(frac=1)"
      ],
      "metadata": {
        "id": "McRygJeDNIlQ",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 424
        },
        "outputId": "e05172b2-6573-448d-a7c4-62652bdef67f"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                               filename  \\\n",
              "1014                              email/2073085598a.tif   \n",
              "1604                                     email/6449.tif   \n",
              "245                              advertisement/3246.tif   \n",
              "2595               scientific _publication/10417856.tif   \n",
              "1056                               email/2085116518.tif   \n",
              "...                                                 ...   \n",
              "2477  scientific _publication/1005092826_1005092867.tif   \n",
              "2653                   scientific _publication/5999.tif   \n",
              "1787                                     email/6911.tif   \n",
              "1851                              email/2085177886b.tif   \n",
              "1212                                     email/1077.tif   \n",
              "\n",
              "                        label  \n",
              "1014                    email  \n",
              "1604                    email  \n",
              "245             advertisement  \n",
              "2595  scientific _publication  \n",
              "1056                    email  \n",
              "...                       ...  \n",
              "2477  scientific _publication  \n",
              "2653  scientific _publication  \n",
              "1787                    email  \n",
              "1851                    email  \n",
              "1212                    email  \n",
              "\n",
              "[2961 rows x 2 columns]"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-149a772c-4180-46a7-b932-3bb22bae8f3b\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>filename</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1014</th>\n",
              "      <td>email/2073085598a.tif</td>\n",
              "      <td>email</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1604</th>\n",
              "      <td>email/6449.tif</td>\n",
              "      <td>email</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>245</th>\n",
              "      <td>advertisement/3246.tif</td>\n",
              "      <td>advertisement</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2595</th>\n",
              "      <td>scientific _publication/10417856.tif</td>\n",
              "      <td>scientific _publication</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1056</th>\n",
              "      <td>email/2085116518.tif</td>\n",
              "      <td>email</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2477</th>\n",
              "      <td>scientific _publication/1005092826_1005092867.tif</td>\n",
              "      <td>scientific _publication</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2653</th>\n",
              "      <td>scientific _publication/5999.tif</td>\n",
              "      <td>scientific _publication</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1787</th>\n",
              "      <td>email/6911.tif</td>\n",
              "      <td>email</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1851</th>\n",
              "      <td>email/2085177886b.tif</td>\n",
              "      <td>email</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1212</th>\n",
              "      <td>email/1077.tif</td>\n",
              "      <td>email</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>2961 rows × 2 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-149a772c-4180-46a7-b932-3bb22bae8f3b')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-149a772c-4180-46a7-b932-3bb22bae8f3b button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-149a772c-4180-46a7-b932-3bb22bae8f3b');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "X = df['filename'].values\n",
        "y = df['label'].values"
      ],
      "metadata": {
        "id": "q2vXYvZJKxeB"
      },
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)\n",
        "X_valid, X_test, y_valid, y_test = train_test_split(X_test, y_test, test_size=0.6, random_state=42)"
      ],
      "metadata": {
        "id": "ku6mg89eJms9"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Criação dos logits dos labels"
      ],
      "metadata": {
        "id": "NmXVHUj6OsbQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.utils.np_utils import to_categorical\n",
        "\n",
        "train_labels = pd.Categorical(y_train)\n",
        "valid_labels = pd.Categorical(y_valid)\n",
        "test_labels = pd.Categorical(y_test)\n",
        "\n",
        "y_train_int = train_labels.codes\n",
        "y_valid_int = valid_labels.codes\n",
        "y_test_int = test_labels.codes\n",
        "\n",
        "y_train_logits = to_categorical(y_train_int)\n",
        "y_valid_logits = to_categorical(y_valid_int)\n",
        "y_test_logits = to_categorical(y_test_int)"
      ],
      "metadata": {
        "id": "8sPRtB2oOujl"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "lista_train = []\n",
        "for i in range(0,len(X_train)):\n",
        "    lista_train.append([X_train[i],y_train[i]])\n",
        "\n",
        "lista_valid = []\n",
        "for i in range(0,len(X_valid)):\n",
        "    lista_valid.append([X_valid[i],y_valid[i]])\n",
        "\n",
        "lista_test = []\n",
        "for i in range(0,len(X_test)):\n",
        "    lista_test.append([X_test[i],y_test[i]])"
      ],
      "metadata": {
        "id": "TkbKFucdQ7C2"
      },
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "df_train = pd.DataFrame(lista_train,columns = ['filename','label'])\n",
        "df_valid = pd.DataFrame(lista_valid,columns = ['filename','label'])\n",
        "df_test = pd.DataFrame(lista_test,columns = ['filename','label'])"
      ],
      "metadata": {
        "id": "agY1B8lPRt0j"
      },
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "datagen = ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "train_generator = datagen.flow_from_dataframe(df_train,\n",
        "                                              directory=path_documents,\n",
        "                                              x_col='filename',y_col='label',\n",
        "                                              class_mode='categorical',\n",
        "                                              shuffle=True)\n",
        "\n",
        "valid_generator = datagen.flow_from_dataframe(df_valid,\n",
        "                                              directory=path_documents,\n",
        "                                              x_col='filename',y_col='label',\n",
        "                                              class_mode='categorical',\n",
        "                                              shuffle=False)\n",
        "\n",
        "test_generator = datagen.flow_from_dataframe(df_test,\n",
        "                                              directory=path_documents,\n",
        "                                              x_col='filename',y_col='label',\n",
        "                                             class_mode='categorical',\n",
        "                                              shuffle=False)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "y9YGsrzzJRrr",
        "outputId": "37d7a872-0164-47a5-9b7c-b54a7276cb6a"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 2072 validated image filenames belonging to 3 classes.\n",
            "Found 355 validated image filenames belonging to 3 classes.\n",
            "Found 534 validated image filenames belonging to 3 classes.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Hyperparameters"
      ],
      "metadata": {
        "id": "A9i8TjKRjHJ1"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "dims = (224,224,3)\n",
        "num_classes = 3\n",
        "lr = 0.00002"
      ],
      "metadata": {
        "id": "vwZ2ynTPjHRI"
      },
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = Sequential()\n",
        "# Parte 1 da AlexNet\n",
        "model.add(layers.experimental.preprocessing.Resizing(224,224,interpolation=\"bilinear\",input_shape=(224,224,3)))\n",
        "model.add(Conv2D(96,(11,11), strides=(4,4)))\n",
        "model.add(Activation(\"relu\"))\n",
        "model.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "\n",
        "# Parte 2 da AlexNet\n",
        "model.add(Conv2D(256,(5,5), strides=(1,1),padding=\"same\"))\n",
        "model.add(Activation(\"relu\"))\n",
        "model.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "\n",
        "# Parte 3 da AlexNet\n",
        "model.add(Conv2D(384,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model.add(Activation(\"relu\"))\n",
        "\n",
        "model.add(Conv2D(384,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model.add(Activation(\"relu\"))\n",
        "\n",
        "model.add(Conv2D(256,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model.add(Activation(\"relu\"))\n",
        "model.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "\n",
        "model.add(Flatten())\n",
        "\n",
        "model.add(Dense(4096))\n",
        "model.add(Activation(\"relu\"))\n",
        "\n",
        "model.add(Dense(4096))\n",
        "model.add(Activation(\"relu\"))\n",
        "\n",
        "model.add(Dense(num_classes))\n",
        "model.add(Activation(\"softmax\"))"
      ],
      "metadata": {
        "id": "ADggsOKeh1Lc"
      },
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model.summary()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1dPoAyxJqk7l",
        "outputId": "e6cf9e4d-8675-4235-ab40-4c3a13b35119"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " resizing (Resizing)         (None, 224, 224, 3)       0         \n",
            "                                                                 \n",
            " conv2d (Conv2D)             (None, 54, 54, 96)        34944     \n",
            "                                                                 \n",
            " activation (Activation)     (None, 54, 54, 96)        0         \n",
            "                                                                 \n",
            " max_pooling2d (MaxPooling2D  (None, 26, 26, 96)       0         \n",
            " )                                                               \n",
            "                                                                 \n",
            " conv2d_1 (Conv2D)           (None, 26, 26, 256)       614656    \n",
            "                                                                 \n",
            " activation_1 (Activation)   (None, 26, 26, 256)       0         \n",
            "                                                                 \n",
            " max_pooling2d_1 (MaxPooling  (None, 12, 12, 256)      0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " conv2d_2 (Conv2D)           (None, 12, 12, 384)       885120    \n",
            "                                                                 \n",
            " activation_2 (Activation)   (None, 12, 12, 384)       0         \n",
            "                                                                 \n",
            " conv2d_3 (Conv2D)           (None, 12, 12, 384)       1327488   \n",
            "                                                                 \n",
            " activation_3 (Activation)   (None, 12, 12, 384)       0         \n",
            "                                                                 \n",
            " conv2d_4 (Conv2D)           (None, 12, 12, 256)       884992    \n",
            "                                                                 \n",
            " activation_4 (Activation)   (None, 12, 12, 256)       0         \n",
            "                                                                 \n",
            " max_pooling2d_2 (MaxPooling  (None, 5, 5, 256)        0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 6400)              0         \n",
            "                                                                 \n",
            " dense (Dense)               (None, 4096)              26218496  \n",
            "                                                                 \n",
            " activation_5 (Activation)   (None, 4096)              0         \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 4096)              16781312  \n",
            "                                                                 \n",
            " activation_6 (Activation)   (None, 4096)              0         \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 3)                 12291     \n",
            "                                                                 \n",
            " activation_7 (Activation)   (None, 3)                 0         \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 46,759,299\n",
            "Trainable params: 46,759,299\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "sgd = SGD(learning_rate=lr)\n",
        "\n",
        "model.compile(optimizer=sgd,loss=\"categorical_crossentropy\",metrics=['accuracy'])\n",
        "\n",
        "history_fine = model.fit(train_generator,\n",
        "                        epochs=50,\n",
        "                        batch_size=64,\n",
        "                        validation_data=valid_generator)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_UPcf4FSHnw6",
        "outputId": "54b793d0-53fe-43dc-ab4d-7bab1265bd11"
      },
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/50\n",
            "65/65 [==============================] - 25s 179ms/step - loss: 1.0992 - accuracy: 0.3277 - val_loss: 1.0997 - val_accuracy: 0.3211\n",
            "Epoch 2/50\n",
            "65/65 [==============================] - 10s 154ms/step - loss: 1.0989 - accuracy: 0.3277 - val_loss: 1.0994 - val_accuracy: 0.3211\n",
            "Epoch 3/50\n",
            "65/65 [==============================] - 10s 153ms/step - loss: 1.0986 - accuracy: 0.3277 - val_loss: 1.0990 - val_accuracy: 0.3211\n",
            "Epoch 4/50\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 1.0982 - accuracy: 0.3277 - val_loss: 1.0987 - val_accuracy: 0.3211\n",
            "Epoch 5/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0979 - accuracy: 0.3277 - val_loss: 1.0984 - val_accuracy: 0.3211\n",
            "Epoch 6/50\n",
            "65/65 [==============================] - 10s 152ms/step - loss: 1.0976 - accuracy: 0.3277 - val_loss: 1.0980 - val_accuracy: 0.3211\n",
            "Epoch 7/50\n",
            "65/65 [==============================] - 10s 152ms/step - loss: 1.0973 - accuracy: 0.3277 - val_loss: 1.0977 - val_accuracy: 0.3211\n",
            "Epoch 8/50\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 1.0970 - accuracy: 0.3277 - val_loss: 1.0974 - val_accuracy: 0.3211\n",
            "Epoch 9/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0966 - accuracy: 0.3277 - val_loss: 1.0971 - val_accuracy: 0.3211\n",
            "Epoch 10/50\n",
            "65/65 [==============================] - 11s 163ms/step - loss: 1.0963 - accuracy: 0.3277 - val_loss: 1.0968 - val_accuracy: 0.3211\n",
            "Epoch 11/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0960 - accuracy: 0.3277 - val_loss: 1.0965 - val_accuracy: 0.3211\n",
            "Epoch 12/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0957 - accuracy: 0.3282 - val_loss: 1.0962 - val_accuracy: 0.3211\n",
            "Epoch 13/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0954 - accuracy: 0.3282 - val_loss: 1.0959 - val_accuracy: 0.3211\n",
            "Epoch 14/50\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 1.0951 - accuracy: 0.3287 - val_loss: 1.0956 - val_accuracy: 0.3211\n",
            "Epoch 15/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0948 - accuracy: 0.3287 - val_loss: 1.0953 - val_accuracy: 0.3211\n",
            "Epoch 16/50\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 1.0945 - accuracy: 0.3287 - val_loss: 1.0950 - val_accuracy: 0.3211\n",
            "Epoch 17/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0942 - accuracy: 0.3287 - val_loss: 1.0947 - val_accuracy: 0.3211\n",
            "Epoch 18/50\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 1.0939 - accuracy: 0.3287 - val_loss: 1.0943 - val_accuracy: 0.3211\n",
            "Epoch 19/50\n",
            "65/65 [==============================] - 11s 163ms/step - loss: 1.0936 - accuracy: 0.3287 - val_loss: 1.0940 - val_accuracy: 0.3211\n",
            "Epoch 20/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0933 - accuracy: 0.3282 - val_loss: 1.0937 - val_accuracy: 0.3211\n",
            "Epoch 21/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0930 - accuracy: 0.3287 - val_loss: 1.0935 - val_accuracy: 0.3211\n",
            "Epoch 22/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0927 - accuracy: 0.3287 - val_loss: 1.0932 - val_accuracy: 0.3211\n",
            "Epoch 23/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0924 - accuracy: 0.3296 - val_loss: 1.0929 - val_accuracy: 0.3211\n",
            "Epoch 24/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0921 - accuracy: 0.3301 - val_loss: 1.0926 - val_accuracy: 0.3211\n",
            "Epoch 25/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0919 - accuracy: 0.3306 - val_loss: 1.0923 - val_accuracy: 0.3268\n",
            "Epoch 26/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0916 - accuracy: 0.3330 - val_loss: 1.0920 - val_accuracy: 0.3268\n",
            "Epoch 27/50\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 1.0913 - accuracy: 0.3349 - val_loss: 1.0918 - val_accuracy: 0.3239\n",
            "Epoch 28/50\n",
            "65/65 [==============================] - 10s 161ms/step - loss: 1.0910 - accuracy: 0.3359 - val_loss: 1.0915 - val_accuracy: 0.3268\n",
            "Epoch 29/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0908 - accuracy: 0.3388 - val_loss: 1.0912 - val_accuracy: 0.3268\n",
            "Epoch 30/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0905 - accuracy: 0.3403 - val_loss: 1.0909 - val_accuracy: 0.3268\n",
            "Epoch 31/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0902 - accuracy: 0.3480 - val_loss: 1.0907 - val_accuracy: 0.3296\n",
            "Epoch 32/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0899 - accuracy: 0.3494 - val_loss: 1.0904 - val_accuracy: 0.3352\n",
            "Epoch 33/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0897 - accuracy: 0.3596 - val_loss: 1.0901 - val_accuracy: 0.3437\n",
            "Epoch 34/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0894 - accuracy: 0.3557 - val_loss: 1.0899 - val_accuracy: 0.3521\n",
            "Epoch 35/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0891 - accuracy: 0.3726 - val_loss: 1.0896 - val_accuracy: 0.3549\n",
            "Epoch 36/50\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 1.0888 - accuracy: 0.3764 - val_loss: 1.0893 - val_accuracy: 0.3577\n",
            "Epoch 37/50\n",
            "65/65 [==============================] - 11s 162ms/step - loss: 1.0886 - accuracy: 0.3769 - val_loss: 1.0890 - val_accuracy: 0.3634\n",
            "Epoch 38/50\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 1.0883 - accuracy: 0.3856 - val_loss: 1.0888 - val_accuracy: 0.3775\n",
            "Epoch 39/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0880 - accuracy: 0.3977 - val_loss: 1.0885 - val_accuracy: 0.3775\n",
            "Epoch 40/50\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 1.0878 - accuracy: 0.4073 - val_loss: 1.0882 - val_accuracy: 0.3944\n",
            "Epoch 41/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0875 - accuracy: 0.4088 - val_loss: 1.0880 - val_accuracy: 0.4056\n",
            "Epoch 42/50\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 1.0872 - accuracy: 0.4165 - val_loss: 1.0877 - val_accuracy: 0.4113\n",
            "Epoch 43/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0869 - accuracy: 0.4262 - val_loss: 1.0875 - val_accuracy: 0.4141\n",
            "Epoch 44/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0867 - accuracy: 0.4392 - val_loss: 1.0872 - val_accuracy: 0.4225\n",
            "Epoch 45/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0864 - accuracy: 0.4411 - val_loss: 1.0869 - val_accuracy: 0.4282\n",
            "Epoch 46/50\n",
            "65/65 [==============================] - 10s 160ms/step - loss: 1.0861 - accuracy: 0.4585 - val_loss: 1.0867 - val_accuracy: 0.4310\n",
            "Epoch 47/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0859 - accuracy: 0.4628 - val_loss: 1.0864 - val_accuracy: 0.4423\n",
            "Epoch 48/50\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 1.0856 - accuracy: 0.4677 - val_loss: 1.0861 - val_accuracy: 0.4507\n",
            "Epoch 49/50\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 1.0853 - accuracy: 0.4855 - val_loss: 1.0859 - val_accuracy: 0.4563\n",
            "Epoch 50/50\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 1.0851 - accuracy: 0.4865 - val_loss: 1.0856 - val_accuracy: 0.4648\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "loss = history_fine.history['loss']\n",
        "val_loss = history_fine.history['val_loss']\n",
        "\n",
        "plt.plot(history_fine.history['loss'])\n",
        "plt.plot(history_fine.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['training set','validation set'], loc='upper left')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "tjI0SUBHYsNk",
        "outputId": "b5120eb4-b234-4bf6-dd9d-d54ee2d9233e"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Analisando-se os resultados finais com as imagens de teste:"
      ],
      "metadata": {
        "id": "EyzKAVDvZavU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "y_prob = model.predict(test_generator)\n",
        "y_pred = np.argmax(y_prob,axis=1)"
      ],
      "metadata": {
        "id": "GOCq_a72Za2h"
      },
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "y_test = test_generator.classes"
      ],
      "metadata": {
        "id": "SLdLzLwTZxjy"
      },
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(y_test,y_pred))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cMTEd_LtaFIX",
        "outputId": "1c45926e-9f41-4579-bc48-afc40258f915"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.93      0.14      0.25       188\n",
            "           1       1.00      0.39      0.56       171\n",
            "           2       0.39      0.99      0.56       175\n",
            "\n",
            "    accuracy                           0.50       534\n",
            "   macro avg       0.78      0.51      0.46       534\n",
            "weighted avg       0.78      0.50      0.45       534\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Bem ruim... tem como melhorar? Vamos adicionar as seguintes camadas Dropout e BatchNormalization:"
      ],
      "metadata": {
        "id": "x_OY4G-I8Fu8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model2 = Sequential()\n",
        "# Parte 1 da AlexNet\n",
        "model2.add(layers.experimental.preprocessing.Resizing(224,224,interpolation=\"bilinear\",input_shape=(224,224,3)))\n",
        "model2.add(Conv2D(96,(11,11), strides=(4,4)))\n",
        "model2.add(Activation(\"relu\"))\n",
        "model2.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "model2.add(BatchNormalization())\n",
        "\n",
        "# Parte 2 da AlexNet\n",
        "model2.add(Conv2D(256,(5,5), strides=(1,1),padding=\"same\"))\n",
        "model2.add(Activation(\"relu\"))\n",
        "model2.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "model2.add(BatchNormalization())\n",
        "\n",
        "# Parte 3 da AlexNet\n",
        "model2.add(Conv2D(384,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model2.add(Activation(\"relu\"))\n",
        "model2.add(BatchNormalization())\n",
        "\n",
        "model2.add(Conv2D(384,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model2.add(Activation(\"relu\"))\n",
        "model2.add(BatchNormalization())\n",
        "\n",
        "model2.add(Conv2D(256,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model2.add(Activation(\"relu\"))\n",
        "model2.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "model2.add(BatchNormalization())\n",
        "\n",
        "model2.add(Flatten())\n",
        "\n",
        "model2.add(Dense(4096))\n",
        "model2.add(Activation(\"relu\"))\n",
        "model2.add(Dropout(0.5))\n",
        "model2.add(BatchNormalization())\n",
        "\n",
        "model2.add(Dense(4096))\n",
        "model2.add(Activation(\"relu\"))\n",
        "model2.add(Dropout(0.5))\n",
        "model2.add(BatchNormalization())\n",
        "\n",
        "model2.add(Dense(num_classes))\n",
        "model2.add(Activation(\"softmax\"))"
      ],
      "metadata": {
        "id": "5h-HiJQU5mKW"
      },
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sgd = SGD(learning_rate=lr)\n",
        "\n",
        "model2.compile(optimizer=sgd,loss=\"categorical_crossentropy\",metrics=['accuracy'])\n",
        "\n",
        "history_fine = model2.fit(train_generator,\n",
        "                        epochs=30,\n",
        "                        batch_size=64,\n",
        "                        validation_data=valid_generator)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qFkWYbSi8XMz",
        "outputId": "18e1057c-8b21-40b9-972e-ecaf2b535e31"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "65/65 [==============================] - 12s 163ms/step - loss: 1.4476 - accuracy: 0.3793 - val_loss: 1.1069 - val_accuracy: 0.3465\n",
            "Epoch 2/30\n",
            "65/65 [==============================] - 9s 146ms/step - loss: 1.2065 - accuracy: 0.4759 - val_loss: 1.1466 - val_accuracy: 0.3465\n",
            "Epoch 3/30\n",
            "65/65 [==============================] - 10s 160ms/step - loss: 1.0725 - accuracy: 0.5502 - val_loss: 1.1793 - val_accuracy: 0.3831\n",
            "Epoch 4/30\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 0.9569 - accuracy: 0.5917 - val_loss: 1.1494 - val_accuracy: 0.5070\n",
            "Epoch 5/30\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 0.9182 - accuracy: 0.6028 - val_loss: 1.1076 - val_accuracy: 0.4789\n",
            "Epoch 6/30\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 0.8510 - accuracy: 0.6501 - val_loss: 0.9445 - val_accuracy: 0.6056\n",
            "Epoch 7/30\n",
            "65/65 [==============================] - 10s 147ms/step - loss: 0.8499 - accuracy: 0.6347 - val_loss: 0.7982 - val_accuracy: 0.6704\n",
            "Epoch 8/30\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 0.7612 - accuracy: 0.6863 - val_loss: 0.6515 - val_accuracy: 0.7380\n",
            "Epoch 9/30\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 0.7787 - accuracy: 0.6771 - val_loss: 0.5690 - val_accuracy: 0.7803\n",
            "Epoch 10/30\n",
            "65/65 [==============================] - 10s 148ms/step - loss: 0.7254 - accuracy: 0.7037 - val_loss: 0.5257 - val_accuracy: 0.8028\n",
            "Epoch 11/30\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 0.7007 - accuracy: 0.7273 - val_loss: 0.4560 - val_accuracy: 0.8169\n",
            "Epoch 12/30\n",
            "65/65 [==============================] - 11s 173ms/step - loss: 0.6741 - accuracy: 0.7350 - val_loss: 0.4450 - val_accuracy: 0.8563\n",
            "Epoch 13/30\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 0.6670 - accuracy: 0.7283 - val_loss: 0.4399 - val_accuracy: 0.8423\n",
            "Epoch 14/30\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 0.6187 - accuracy: 0.7597 - val_loss: 0.4296 - val_accuracy: 0.8563\n",
            "Epoch 15/30\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 0.6726 - accuracy: 0.7457 - val_loss: 0.4246 - val_accuracy: 0.8394\n",
            "Epoch 16/30\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 0.6430 - accuracy: 0.7577 - val_loss: 0.4277 - val_accuracy: 0.8225\n",
            "Epoch 17/30\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 0.6348 - accuracy: 0.7587 - val_loss: 0.4133 - val_accuracy: 0.8507\n",
            "Epoch 18/30\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 0.6247 - accuracy: 0.7630 - val_loss: 0.3962 - val_accuracy: 0.8563\n",
            "Epoch 19/30\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 0.5889 - accuracy: 0.7698 - val_loss: 0.4136 - val_accuracy: 0.8507\n",
            "Epoch 20/30\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 0.5978 - accuracy: 0.7732 - val_loss: 0.4068 - val_accuracy: 0.8620\n",
            "Epoch 21/30\n",
            "65/65 [==============================] - 10s 160ms/step - loss: 0.5917 - accuracy: 0.7736 - val_loss: 0.3975 - val_accuracy: 0.8592\n",
            "Epoch 22/30\n",
            "65/65 [==============================] - 10s 149ms/step - loss: 0.5705 - accuracy: 0.7722 - val_loss: 0.4038 - val_accuracy: 0.8479\n",
            "Epoch 23/30\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 0.5858 - accuracy: 0.7770 - val_loss: 0.4051 - val_accuracy: 0.8845\n",
            "Epoch 24/30\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 0.5894 - accuracy: 0.7809 - val_loss: 0.3824 - val_accuracy: 0.8958\n",
            "Epoch 25/30\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 0.5848 - accuracy: 0.7847 - val_loss: 0.3874 - val_accuracy: 0.8620\n",
            "Epoch 26/30\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 0.5416 - accuracy: 0.7968 - val_loss: 0.3668 - val_accuracy: 0.8648\n",
            "Epoch 27/30\n",
            "65/65 [==============================] - 10s 150ms/step - loss: 0.5623 - accuracy: 0.7896 - val_loss: 0.3890 - val_accuracy: 0.8592\n",
            "Epoch 28/30\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 0.5488 - accuracy: 0.7939 - val_loss: 0.3796 - val_accuracy: 0.8648\n",
            "Epoch 29/30\n",
            "65/65 [==============================] - 10s 151ms/step - loss: 0.5275 - accuracy: 0.7944 - val_loss: 0.3720 - val_accuracy: 0.8676\n",
            "Epoch 30/30\n",
            "65/65 [==============================] - 10s 158ms/step - loss: 0.5511 - accuracy: 0.7905 - val_loss: 0.3685 - val_accuracy: 0.8704\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Agora avaliando-se o modelo 2 com os dados de teste"
      ],
      "metadata": {
        "id": "WmY7ogEv8cJq"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "y_prob = model2.predict(test_generator)\n",
        "y_pred = np.argmax(y_prob,axis=1)\n",
        "y_test = test_generator.classes\n",
        "from sklearn.metrics import classification_report\n",
        "print(classification_report(y_test,y_pred))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "17lJMon38cTa",
        "outputId": "539601a9-ee11-4b91-9791-74dda3366d6f"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.89      0.80      0.84       188\n",
            "           1       0.89      0.92      0.90       171\n",
            "           2       0.80      0.86      0.83       175\n",
            "\n",
            "    accuracy                           0.86       534\n",
            "   macro avg       0.86      0.86      0.86       534\n",
            "weighted avg       0.86      0.86      0.86       534\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "**Dicas de como ajustar os hiperparâmetros**:\n",
        "\n",
        "*   Número de épocas: \n",
        "*   Batch size:\n",
        "*   \n",
        "\n"
      ],
      "metadata": {
        "id": "-s63L-VQRid7"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Melhorou um pouco!"
      ],
      "metadata": {
        "id": "y-7Ys77AcXTC"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### Data Augmentation\n",
        "\n",
        "Uma estratégia bem conhecida para aumentar a quantidade de dados de treinamento é a data augmentation. Nesse notebook, iremos abordar o uso do [ImageDataGenerator](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator), que pega uma imagem do treinamento e gera novas imagens após a aplicação de várias transformações afins e geométricas (veja a lista de transformações disponíveis na própria documentação).\n",
        "\n",
        "Observação 1: De acordo com a documentação do ImageDataGenerator, as imagens originais não são incluídas nesse novo conjunto de imagens transformadas, nem são aproveitadas no treinamento.\n",
        "\n",
        "Observação 2: **deve-se realizar data augmentation apenas nos dados de treinamento**! Os dados de validação e de teste devem permanecer conforme a amostragem realizada, uma vez que eles devem possuir a mesma distribuição estatística como ocorrem naturalmente.\n",
        "\n",
        "Para isso, propomos a seguinte alteração nos objetos ImageDataGenerator:\n",
        "\n",
        "*   train_datagen: um ImageDataGenerator específico para receber os dados de treinamento e que contém as transformações para a geração de novas imagens;\n",
        "*   agnostic_datagen: um ImageDataGenerator que apenas recebe as imagens de validação e de treinamento, mas que serão armazenadas em fluxos distintos.\n",
        "\n"
      ],
      "metadata": {
        "id": "rruH9aVLkJO_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "agnostic_datagen = ImageDataGenerator(rescale=1./255)\n",
        "\n",
        "valid_generator = agnostic_datagen.flow_from_dataframe(df_valid,\n",
        "                                              directory=path_documents,\n",
        "                                              x_col='filename',y_col='label',\n",
        "                                              class_mode='categorical',shuffle=False)\n",
        "\n",
        "test_generator = agnostic_datagen.flow_from_dataframe(df_test,\n",
        "                                              directory=path_documents,\n",
        "                                              x_col='filename',y_col='label',\n",
        "                                             class_mode='categorical',shuffle=False)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "a5E6vp4ckUvO",
        "outputId": "388e11fa-b0b5-4958-af85-ce3483d29e5a"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 355 validated image filenames belonging to 3 classes.\n",
            "Found 534 validated image filenames belonging to 3 classes.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_datagen = ImageDataGenerator(rescale=1./255,\n",
        "                                   width_shift_range=0.2,\n",
        "                                   height_shift_range=0.2,\n",
        "                                   zoom_range=[0.3, 1.5],\n",
        "                                   brightness_range=[0.1, 1.5]\n",
        "                                   )\n",
        "\n",
        "train_generator = train_datagen.flow_from_dataframe(df_train,\n",
        "                                              directory=path_documents,\n",
        "                                              x_col='filename',y_col='label',\n",
        "                                              class_mode='categorical',\n",
        "                                              shuffle=True)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tBG0-OOkMHm5",
        "outputId": "fb25ca8a-eef4-476e-8b5e-a44e68e1a2ad"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 2072 validated image filenames belonging to 3 classes.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Vamos ver como fica"
      ],
      "metadata": {
        "id": "Xkxh5JMzN1XL"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "img = image.load_img('/content/documents/advertisement/0000126140.tif', target_size= (500,500))\n",
        "img_tensor = image.img_to_array(img)\n",
        "img_tensor = np.expand_dims(img_tensor, axis=0)\n",
        "\n",
        "pic = train_datagen.flow(img_tensor, batch_size =1)\n",
        "plt.figure(figsize=(10,8))\n",
        "#Plots our figures\n",
        "for i in range(1,4):\n",
        "  plt.subplot(1, 3, i)\n",
        "  batch = pic.next()\n",
        "  img = 255*batch[0]\n",
        "  image_ = img.astype('uint8')\n",
        "  plt.imshow(image_)\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 216
        },
        "id": "TfqiB3w9N0Ll",
        "outputId": "ce6c3a38-236f-48f3-bab3-04029878c378"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 720x576 with 3 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model_aug = Sequential()\n",
        "# Parte 1 da AlexNet\n",
        "model_aug.add(layers.experimental.preprocessing.Resizing(224,224,interpolation=\"bilinear\",input_shape=(224,224,3)))\n",
        "model_aug.add(Conv2D(96,(11,11), strides=(4,4)))\n",
        "model_aug.add(Activation(\"relu\"))\n",
        "model_aug.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "model_aug.add(BatchNormalization())\n",
        "\n",
        "# Parte 2 da AlexNet\n",
        "model_aug.add(Conv2D(256,(5,5), strides=(1,1),padding=\"same\"))\n",
        "model_aug.add(Activation(\"relu\"))\n",
        "model_aug.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "model_aug.add(BatchNormalization())\n",
        "\n",
        "# Parte 3 da AlexNet\n",
        "model_aug.add(Conv2D(384,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model_aug.add(Activation(\"relu\"))\n",
        "\n",
        "model_aug.add(Conv2D(384,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model_aug.add(Activation(\"relu\"))\n",
        "\n",
        "model_aug.add(Conv2D(256,(3,3), strides=(1,1),padding=\"same\"))\n",
        "model_aug.add(Activation(\"relu\"))\n",
        "\n",
        "model_aug.add(MaxPooling2D((3, 3),strides=(2,2)))\n",
        "\n",
        "model_aug.add(Flatten())\n",
        "\n",
        "model_aug.add(Dense(4096))\n",
        "model_aug.add(Activation(\"relu\"))\n",
        "model_aug.add(Dropout(0.5))\n",
        "model_aug.add(BatchNormalization())\n",
        "\n",
        "model_aug.add(Dense(4096))\n",
        "model_aug.add(Activation(\"relu\"))\n",
        "model_aug.add(Dropout(0.5))\n",
        "model_aug.add(BatchNormalization())\n",
        "\n",
        "model_aug.add(Dense(num_classes))\n",
        "model_aug.add(Activation(\"softmax\"))"
      ],
      "metadata": {
        "id": "5vxlly1tL_Dh"
      },
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sgd = SGD(learning_rate=lr)\n",
        "\n",
        "model_aug.compile(optimizer=sgd,loss=\"categorical_crossentropy\",metrics=['accuracy'])\n",
        "\n",
        "history_fine = model_aug.fit(train_generator,\n",
        "                        epochs=30,\n",
        "                        batch_size=64,\n",
        "                        validation_data=valid_generator)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jIyuxL00NIu2",
        "outputId": "9251d64a-28ee-4e79-d807-e8a80f098672"
      },
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/30\n",
            "65/65 [==============================] - 39s 591ms/step - loss: 1.4015 - accuracy: 0.4083 - val_loss: 1.0997 - val_accuracy: 0.3465\n",
            "Epoch 2/30\n",
            "65/65 [==============================] - 39s 606ms/step - loss: 1.3107 - accuracy: 0.4319 - val_loss: 1.1185 - val_accuracy: 0.3465\n",
            "Epoch 3/30\n",
            "65/65 [==============================] - 38s 579ms/step - loss: 1.1424 - accuracy: 0.5048 - val_loss: 1.1324 - val_accuracy: 0.3465\n",
            "Epoch 4/30\n",
            "65/65 [==============================] - 37s 576ms/step - loss: 1.1521 - accuracy: 0.5039 - val_loss: 1.1286 - val_accuracy: 0.3493\n",
            "Epoch 5/30\n",
            "65/65 [==============================] - 39s 603ms/step - loss: 1.1056 - accuracy: 0.5232 - val_loss: 1.0799 - val_accuracy: 0.3690\n",
            "Epoch 6/30\n",
            "65/65 [==============================] - 37s 574ms/step - loss: 1.0647 - accuracy: 0.5526 - val_loss: 0.9405 - val_accuracy: 0.4817\n",
            "Epoch 7/30\n",
            "65/65 [==============================] - 39s 605ms/step - loss: 1.0613 - accuracy: 0.5555 - val_loss: 0.8024 - val_accuracy: 0.6732\n",
            "Epoch 8/30\n",
            "65/65 [==============================] - 38s 584ms/step - loss: 1.0061 - accuracy: 0.5830 - val_loss: 0.6759 - val_accuracy: 0.7127\n",
            "Epoch 9/30\n",
            "65/65 [==============================] - 38s 581ms/step - loss: 1.0416 - accuracy: 0.5656 - val_loss: 0.5606 - val_accuracy: 0.7887\n",
            "Epoch 10/30\n",
            "65/65 [==============================] - 39s 602ms/step - loss: 0.9426 - accuracy: 0.6052 - val_loss: 0.5411 - val_accuracy: 0.7887\n",
            "Epoch 11/30\n",
            "65/65 [==============================] - 38s 578ms/step - loss: 0.9784 - accuracy: 0.6028 - val_loss: 0.5608 - val_accuracy: 0.7859\n",
            "Epoch 12/30\n",
            "65/65 [==============================] - 39s 600ms/step - loss: 0.9338 - accuracy: 0.6221 - val_loss: 0.5427 - val_accuracy: 0.7859\n",
            "Epoch 13/30\n",
            "65/65 [==============================] - 38s 582ms/step - loss: 0.9357 - accuracy: 0.6255 - val_loss: 0.5386 - val_accuracy: 0.7887\n",
            "Epoch 14/30\n",
            "65/65 [==============================] - 39s 599ms/step - loss: 0.9314 - accuracy: 0.6226 - val_loss: 0.5919 - val_accuracy: 0.7662\n",
            "Epoch 15/30\n",
            "65/65 [==============================] - 38s 587ms/step - loss: 0.8912 - accuracy: 0.6390 - val_loss: 0.6235 - val_accuracy: 0.7634\n",
            "Epoch 16/30\n",
            "65/65 [==============================] - 38s 578ms/step - loss: 0.9057 - accuracy: 0.6351 - val_loss: 0.6185 - val_accuracy: 0.7690\n",
            "Epoch 17/30\n",
            "65/65 [==============================] - 39s 600ms/step - loss: 0.9024 - accuracy: 0.6486 - val_loss: 0.6107 - val_accuracy: 0.7690\n",
            "Epoch 18/30\n",
            "65/65 [==============================] - 38s 579ms/step - loss: 0.8756 - accuracy: 0.6670 - val_loss: 0.5790 - val_accuracy: 0.7606\n",
            "Epoch 19/30\n",
            "65/65 [==============================] - 39s 602ms/step - loss: 0.8624 - accuracy: 0.6713 - val_loss: 0.5704 - val_accuracy: 0.7775\n",
            "Epoch 20/30\n",
            "65/65 [==============================] - 38s 580ms/step - loss: 0.9377 - accuracy: 0.6289 - val_loss: 0.5641 - val_accuracy: 0.7746\n",
            "Epoch 21/30\n",
            "65/65 [==============================] - 38s 580ms/step - loss: 0.8503 - accuracy: 0.6718 - val_loss: 0.5709 - val_accuracy: 0.7887\n",
            "Epoch 22/30\n",
            "65/65 [==============================] - 39s 599ms/step - loss: 0.8596 - accuracy: 0.6641 - val_loss: 0.5945 - val_accuracy: 0.7746\n",
            "Epoch 23/30\n",
            "65/65 [==============================] - 38s 577ms/step - loss: 0.8618 - accuracy: 0.6723 - val_loss: 0.5606 - val_accuracy: 0.7944\n",
            "Epoch 24/30\n",
            "65/65 [==============================] - 39s 599ms/step - loss: 0.8790 - accuracy: 0.6684 - val_loss: 0.6085 - val_accuracy: 0.7746\n",
            "Epoch 25/30\n",
            "65/65 [==============================] - 38s 577ms/step - loss: 0.8469 - accuracy: 0.6665 - val_loss: 0.5702 - val_accuracy: 0.7915\n",
            "Epoch 26/30\n",
            "65/65 [==============================] - 39s 599ms/step - loss: 0.8401 - accuracy: 0.6747 - val_loss: 0.5139 - val_accuracy: 0.7887\n",
            "Epoch 27/30\n",
            "65/65 [==============================] - 37s 575ms/step - loss: 0.8048 - accuracy: 0.6998 - val_loss: 0.6202 - val_accuracy: 0.7775\n",
            "Epoch 28/30\n",
            "65/65 [==============================] - 38s 579ms/step - loss: 0.8119 - accuracy: 0.6839 - val_loss: 0.6031 - val_accuracy: 0.7859\n",
            "Epoch 29/30\n",
            "65/65 [==============================] - 39s 599ms/step - loss: 0.8182 - accuracy: 0.6858 - val_loss: 0.6779 - val_accuracy: 0.7465\n",
            "Epoch 30/30\n",
            "65/65 [==============================] - 38s 577ms/step - loss: 0.8082 - accuracy: 0.6940 - val_loss: 0.5469 - val_accuracy: 0.7944\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "y_prob = model_aug.predict(test_generator)\n",
        "y_pred = np.argmax(y_prob,axis=1)\n",
        "y_test = test_generator.classes\n",
        "print(classification_report(y_test,y_pred))"
      ],
      "metadata": {
        "id": "WVu0O1MTTL8L",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f87fbf26-553f-4050-8d52-541876dcf036"
      },
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.88      0.77      0.82       188\n",
            "           1       0.90      0.80      0.85       171\n",
            "           2       0.70      0.87      0.78       175\n",
            "\n",
            "    accuracy                           0.81       534\n",
            "   macro avg       0.83      0.81      0.82       534\n",
            "weighted avg       0.83      0.81      0.82       534\n",
            "\n"
          ]
        }
      ]
    }
  ]
}