1-CNN-Teoria.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Redes Neuronales Convolucionales (CNN)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![cnn.jpeg](imgs/cnn.jpeg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- La meta es reducir el ancho/alto y aumentar la profundidad\n",
    "- Transformar información espacial en features (feature maps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "from IPython.display import YouTubeVideo"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Videos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Intro CNNs:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/html": [
       "\n",
       "        <iframe\n",
       "            width=\"400\"\n",
       "            height=\"300\"\n",
       "            src=\"https://www.youtube.com/embed/HrYNL_1SV2Y\"\n",
       "            frameborder=\"0\"\n",
       "            allowfullscreen\n",
       "        ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<IPython.lib.display.YouTubeVideo at 0x1070d90b8>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "YouTubeVideo('HrYNL_1SV2Y')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Capas convolucionales"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/html": [
       "\n",
       "        <iframe\n",
       "            width=\"400\"\n",
       "            height=\"300\"\n",
       "            src=\"https://www.youtube.com/embed/h5R_JvdUrUI\"\n",
       "            frameborder=\"0\"\n",
       "            allowfullscreen\n",
       "        ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<IPython.lib.display.YouTubeVideo at 0x10627e278>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "YouTubeVideo('h5R_JvdUrUI')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/html": [
       "\n",
       "        <iframe\n",
       "            width=\"400\"\n",
       "            height=\"300\"\n",
       "            src=\"https://www.youtube.com/embed/RnM1D-XI--8\"\n",
       "            frameborder=\"0\"\n",
       "            allowfullscreen\n",
       "        ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<IPython.lib.display.YouTubeVideo at 0x62ceb3358>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "YouTubeVideo('RnM1D-XI--8')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Capas pooling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wCEABALDA4MChAODQ4SERATGCgaGBYWGDEjJR0oOjM9PDkzODdASFxOQERXRTc4UG1RV19iZ2hnPk1xeXBkeFxlZ2MBERISGBUYLxoaL2NCOEJjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY2NjY//AABEIAWgB4AMBIgACEQEDEQH/xAAbAAEBAQEBAQEBAAAAAAAAAAAABQQDAgYBB//EAD0QAAICAQIDBwIEBQMCBgMAAAABAgMRBBIFITETFUFRUpHRImEUIzJxM3KBobFCYsEkNFSCkqLh8AZDY//EABkBAQADAQEAAAAAAAAAAAAAAAABAgMEBf/EACoRAQACAgEEAgIBAwUAAAAAAAABAgMREgQTITEiQSOBoUJRkRQyYXGx/9oADAMBAAIRAxEAPwD+fgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA1932+qHu/gd32+qHu/gnSNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNwyA1932+qHu/gd32+qHu/gaNsgNfd9vqh7v4Hd9vqh7v4GjbIDX3fb6oe7+B3fb6oe7+Bo2yA1932+qHu/gd32+qHu/gaNsgNfd9vqh7v4Hd9vqh7v4GjbIDX3fb6oe7+B3fb6oe7+Bo2yA1932+qHu/gd32+qHu/gaNsgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcMgNfd9vqh7v4Hd9vqh7v4GjcKQALqAAAAAAAAAAAAAAAAAAAAAAAAAAABdQF1A1X2qu+cI01Yi8LMT3VG+6O6vS1teew52xU+IuL6OzDL1qddP5b27fJG1azaZcWXJGOtfHmUGyydUttmnqi/vA8/iG//wBNP/oK/Fq4y0bm19UXyZP4RXGeqblz2rKRW1Zi3FfHkpfF3NCq1LhuWjrx/KcHe08OipNf7C9ZKyNif1YcopYxjHiTONVxjdCaWHJcy16TWN7Z4c1cltTDLC2dktsNPVJ+SgdbIaiqO6ekrS89hs4PCMdNK3GZNmymTnHE8tyipc8Y5k1pNo3tGXPWlpiK+kD8T/8Axp/9B1pdt7/K01Uv2gctZCNershHonyLlajp9LXGHLK64/uUpE2ltmvXHWJiN7SLXbQ/zdNVH94H5VYrXKMqql9EnlR+xYsjHUaD6+eYZz98EPS/xJfyS/wTaJrKuO8ZKTOtTDiADJ1gAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABdQF1A7aptayxrk1IpU8Wrde2+Lz44WUybq/8Au7f5jXTwmydalZNQz4YNKzaLTxc9645xx3HjX8Q/ErZBOME88/Ezaa+WnuVkfDqvM6avRWaVpyxKL6SRxqrlbYoQWZMrMzvz7aUikU1X0rriuma3uuW/HkTNXqZaq7fJYXRLyNq4NLbztW7ywT76Z0WOuxYaLXm/9TLDXDEz2/bRoNc9K2pLdB+HkbJ8UohB9hW9z+2CdpdJZqp4hyS6tmuzhE4wbrsU2vDArN9eDJXDN/n7T5zc5uUnlt5ZS0XFFXWq7k2l0kiY002n1Rs0nDbNRDe5KEPBvxK1m2/i0yxSa/k9O2s4nGyp1URaTWG35GLS/wASX8kv8HfV8Os08N6kpw8WvA4aX+JL+SX+CZmZt8laxSMf4/TiACjcAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAALqAuoGqxpcSbl07Tn7ly+MZKOXyTyvp3Hz+r/AO6t/mO1HE76YbOUkumTWl4rM7ceXDbJWs1n0o8Ra7u/MSUnjl5MxcGcVqnnq48jLqdVbqZJ2PkuiXRHKEpQkpReGujItfduTTHh44pxzPt9HJTeohJxzFPCeeiwydxtx7WtL9STyc1xbUKGMRb88GKyyVs3ObzJ+Ja+SLRqGWDprY7crLPBnF6Vpfq3czXVBKc5qKjnksLHJHz1Gos0891csea8zRbxTUWQ2rEc+KJpliK6lTN0lr3m0T4lx1zi9Za49MlzlLR17FmOI9FnH9D5w06bXXaZbYtOPkylL8Z8ujPhnJWIj6WniOin2iwsPwxn+hC0v8SX8kv8HvU667UrbJpR8keNL/El/JL/AALW5WjSuPFOPHO/cuIAM3UAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAF1AXUDvqYuWtnFdXLCKdfDtPTXHtlKc35E62fZ8QlP0zyXW431xlBRmnz/AFYwbY6xMztw9Re9KV4+kriOgjRBW1Z2+KfgZdHp3qb1WnhdW/sU+J2qrRqmUt1ksJ/Ji4VdGnVfW8KSxkraK89fTXFfJOHlPtRXD9JnstrcsZzkla7SvS3bc5i+aZ9BjNinywotf4I3GLo2XRhB52LmzTLWsR4c/SZclr6t5h54doVqc2WN7FywvE3T4dpbYuNcZQkl90c+D3x7J05Sknlfc3Q3VqU7tsX4tSFK1mvlXqMuWuSYj9PnLa3VZKEusXgp6Th1UaFbqMycukUT9XartTZZHo3yLmntjqdLFx5tJcs4wzPHETPl09Te9aRNf2x6vh1TodunynHrEn6X+JL+SX+C3fZHS6Se9/U88s9WyJpf4kv5Jf4JvERbwrhva2KeTiADJ2AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABdQF1A7av8A7q3+Y81ztisVSmv5TpdDtNfKHqnguJV6WMK4Risp839jStOUy5r5oxUr43t85Jycm5Nt/c/C1xbTQlp+2ikpR8V4ow8M08dRqcTWYxWWvMrNJi3Fematsfc+nBWX7MKdmz93g5H0zzGcYKENr/wSOLaeNN0ZQWFNdC18c1jbPD1NcluOtMKbTysp/Y92WXSilZKbX+5spcJ08OylfNJvwz4G9bNTW1OMWmunkTXFNo2jJ1Vcdta9Pmj3XKyMvy3JP/aetTV2OonWukXyLelojpdPDbFOcsZb8ylaTadNcuaMddz9oNkpyl+Y5N/7jppf4kv5Jf4LWqohqtNJyilOOcP7oi6X+JL+SX+C01mtla5Yy45mHEAGboAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAD9jFy/ZeLA/Bh+R73KP6eb82XdFoabdP2lm6cpt89z5cwyzZq4q8rPn8PyGH5FPRaWFnE7KJ2SlCGcc+p74xpq9Mq51SlHc8OO4I79O5GP7lKw/I/MPyPo+7aPwmMy3bc79zMXCdJG/tJXScnHlt3dApHVY5ra31CSF1KHF6Y0WqEJNx5NJvoTw3raL1i0fbvqJOGtnJdVPKK1et098Iylb2U0sPJKnfVOblKjLfN/Wz3XXG2OYaRtee9mkW4zvbG2LuRFZidx/00cR1tc6VRQ8x8WZNDqfwuoU2sxfJn7N11y2z0ri/vNnntKP8Aw/8A72RM7ne1q0iteHHx+lpazSv8ztl0xjPQk8Q1S1Vycf0R5I9KjMd34J4/nZxdlKeHpsP+dk2vN/G1MeCMU8orP8NXDNbCmMqbuUJdH5G2Ws01EXJXdo8YSXMkwlVOW2Glbfkps6WVKuO6WjaXnvYjJNY1tF+nrknlNZ/hmusd1srH1k8lbSa+m2iMLp9nOOOZN7Wj/wAP/wC9nuqMLf4ekcv/ADsiJ4+dtMlIyRxms/w36vXU1aeVdM985Z5/uTNL/El/JL/B0sUKv4mkcf8Azs8q+qKeyjDaazub6ibcp3Mq1x8KzWIn+GcAFHQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAASbeEelDlmTwv8nWj6rNkPpzyz49SEx5c8Rj+rm/JHmUnLr08inqNKoUSlCc1KKzzl1POj06toVlk5ty6Yl0M+7XjydH+lyc+H2mlTTaXXWV2S09mytt8nLGThCiT1sqXbLaufXmVfxUdLp+dbnh4RbuV3EOfNgzdqb0jepRdNRqJ6vs6sxti3l5xg66/T6muxPVWKXlLOSjp1ssdzeZT5tZ5LPkYeJ638Q41RhtjB55+LFbxbelcmLNjvWbR4mHV6XX/gfpk1XjOzdzwZuHUaq2yT0s9mOss4Nnff8A023sn2uMZ8DNw3iC0e+M4OUZc+XVMu5YnNwtusb+nHWVXVTlHUPM92c5zkzGzX6v8ZPtNu1J4SMYdNN8Y5e3umKndCL6OSTLU00oqOVFddpDTaaa6opQ19M4JXRw19sowzVtbWnodHlpTcW8betXFWaFTkvqSTM3DYRle5P/AErKGs1iujsrTUfFvxOGnulRaprn4NeZNK2imlMuSk5otHmPH7WG32sefJp8jBxSCU4TXWS5nfvCjGXuz5YMGq1D1Fm7GEuSRnhpatty6erz4701Wdy3cNglp3P/AFSfU0V5zL6m49OZN0er7DMZJuD8vA1T4hTGH5abfgsYIyY7zbwt0/UYq4oi31/LDq4KvUzjHomVUlTRCEOXLrj+5GnJzm5S6t5Zu02vjGtQuzy5Jo1y1maxpy9LkpTJMz421OKu0uJ88x6/cilDU66DrddKfNYzjGCeMNZrHlHV5K3vHEABs5AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAfsYuXRHrMYdPqfn4Gzh2m/E8nNxgubx4jXaZ6a2Crm3GfJZxyLcZ1yZ92nPhvywttvLeWeqs5lt645Y/cq3cNUNPKUbZb4rPNLDPHC9N20XdObwnhJE9u29Kx1GPjN9+IZ9R+M7D839HjjH9zzpPxfZvsP0/fBRvcXf+FT+qfjjojpOqGk0+dzcY9c9TDhfhM8fTunNgjNWO5PmPaPCN34jOZdrnourPodMq3p/zFHP+rJBhq0tbK5x+l8is9Etbpl+ZsTeVyI+XKPDDqa4p6e3PJMef8uehT/FT3P/AKfn2ef7f0OPHFQuz27e0y848jtpX2lr075WVrEvLkYeKaF6aSsUt0ZvH7Mmm/O40pnjF3KTXJMzr0rbdH3f0h2Wzr/98TDwJUN2b9rs8N3kO5Jfht3a/mYztxyM3DeHvWbpOeyMeXLq2aOCIx9u/wA/H/j3xhUrUPsceG7HTJONeu0j0c+zctybymZA7sWuEanYba+HSlBSsmoZ8DNp8K+vd03LJZtSaWc8n4RyY5bzXUQ7+lwVy7m30k6nSz0+G3ui+jRyrrlbNQgstlXWP/oXvSTaXL7mXhbXbyz128hW8zTkjJgrXNGOJ8S9d2vHO1bvLBjuqlTY4TXMr4krm8N5fl4Y8zJxVrdWv9WGUx5LWtqWvU9NSlOVf7s2n009RJqPJLq2aJ8Okot12KTXhg78Nx+F5dcvJ2rX5kmlyx6cEXy2i0xC+DpKXxxafcojWHh9UatPoZ3Q3ykoRfTPieNa09XZt6ZKjw6YbVmPJ4XkaZLzWImHP0+GuS8xb6TtRop0w3xkpx8ceBlLawqJ7liPN4fkRCcV5tHlHVYa4rRx+wAGrlADvotM9Xq6qIyUXOWMvwXiwOAKE6NLOvdp9Pq3BTUe2m1tfPxwuXue7uFN6q6NdlVVaulVUrJYcmn0X9uZCUwFRcJ36TT4nXVfOc4ONksOUk0kkZqNBZfCe2ypWRUn2Tb38uvgBkBp19MKLKlWmlKmE3+7imzXqOE7VpoVWZtnJVWqfJRsaTS9nj90wJYNlHDb7oKS2QW6UXveMbVmTf2WT2uE3OcvzaVWq1b2jl9LjnGenmBgBrr0ErarJ130ycFKTgpPc0ur6HSXCbo1Tl2tTnDbuqTe5bui6YAwA16rh1ulrdjnVYoz2T2Szsl5MyEoAAAAAAAAAAAAAAAAAAAAAAAAatJLURlH8Mm5POV9uR+a2eolcvxKaklyXgduG6qvTP8AMziWVleA4jrK77q3WsqHPL8S/jj7Ybt3f9vjXt+3W8Qek/Mi1W1zljng/OFy1Cm41JuH+rlyRrv4nRPTSUcuUo4246HLhOrrrg6Zpp5yml1LTEco+TKJv25/H536/u7aqHJ2JYuX6WlzydI9tZp4rUxWX1WDlqN0NWtVnNcOsPt5na62Oq02KZNKXic0RXtz8/t6drZP9RT8P1/hMrppfEZV8nBdF9/Io3ayWjoSht5vCTWcEiOmsep7JPEl4nvWaaypKcrHYumX4Ff64+S81i+G1bY9xv8Awp6ZqCV0Wt0lmU/Mn6/Vy1U0pzxGPRKJ+16K6VHKzGVnYZLVieH5L/BOP3PnavUVieEzTU6bu9rvw/Zb10xu288e5y0WsnpJPspblLrFx/8AkyqHLMnhf5DnhYgsL+7NXH2qamNe3fW6iWonusknLyS6IzABeIiI1Aa6uIW1xUZJTS6ZMsYuUlFdW8IqQ0dFNadi3PxZnktWI+TpwUyWmZxzpgv1Nmoa3vkuiRyhKUJKUXhrozbrdHGqHaV/p8V5GfTUPUW7M4XVsmtqzXcelL0vF+NvbuuJW7cbI58zJZZK2bnN5bKn4TTL6HW+fLPMwavT/h7MJ5i+aZTHakz8W2fHmrWJyTuHmi+dEswfXqn4neziNso4iox+6Gi0iuTnZ+hckvM0vR0Ww+iDh5MXvSLeTFiz2p8PSUaKNZZRHasSj5PwONkHXZKEusXgoabQ1qtTuWW1nGeSLXtWI+TLDTJa+sftlv1ll8drxGPkjOUb9HXOl2ULDSzjzJxNJrMfEzVvW3z9gALsQ66XUT0uprvrxureVnozkAKXeGmrpuhp9NZX223Kd2YxxJS5LH2P2XFKrbHLUaaU9t8rq1GzGMvOHy5rl9jtwzR1W8PhbLSwubvcJylY47Y4XPqvNnHuyiVsYrVbVfZKOnW3O5J4Tb8Mv9yEv2PFq5Oqeo0znbVbK2Mo2bU23nDWOnI6Vcc2VJSpnv2zjLbZthJyz9Tjjrz8zjVwrc6qrL9mpvTdde3K6tc34ZaZ0t0db00ZQhCL/CVzbxzy54yDyyS1kZ6vT3Tp3RphCLhu/VtWP74O0+Maq+q6GonK1TalD6sdnJPKa/we7uExjO6unVKyymyNdicNq5vHJ/uctdw+Gl0/aK2yUu1dW2VW3mur69AO8uOSetjfGlwgq5QlCE8NuTzKSeOTz/g4X8R7SN8UrpK2tQzbbvaxLPkc9bTXVTo5Qjh2Ub5fd7pL/hGuPBd1dU+1nBTnGtu2px5yTxjnzWVgD8r4xGGkVCqmvyXS8WYhzz9W3HU263UVqhuNyXYKt1SjdGStlHCWYpZ83zfIn6Xg9t8YOUnFtTk4KGZKMWl088vB6s4OqZWyv1DrqrhGeXX9TUnjGM8nkDnxHiX42PL8Qsy3OM7t8V+ywTz9eMvHNeB+EoAAAAAAAAAAAAAAAAAAAAAAAAUeF0VXy/NW7blqL8RxPTU1X19niG/ql4fc4aOi6+UVTLY45blnofmt091NyVsnNy6Sz1L/ANHpz6/N/u+vSpqNDpo6SWIKLjHKl4nDhGnqlXK2SUpp4/Y5XaHVx0uZW7oxWXDPQ56DS6i5SnTZ2cemc9S+/lHxYxH4rfk+/bdfZXLWLS5SjL9XP+x21Cr02nzGKil0RGlprlrOxfO1vrn+5p1Wl1NUVO23tIpNdenIzileMxxdM5ck5aW7nnX+XCOocdS798W31WH0PWq1S1EVFOMYp58eZjPe1R/X19KM+Mb26e7fjNd+Jbqta+yUVtcksbuZjsnHe2vqfm+h4lJvl0XkjyIrFfRfJa+otPp+ttvLeWfgBZmAAD3TPs7oTfSLTLUsWVfS01JcmQjpXZbDlXKS+yMsmPm6en6js7jW4lS4hbGGncM/VPojJw62Nd7UnjcsZMspSlLMm2/ufhNccRXijJ1Fr5e4u4m7ctLaunMn8TtjKyMIvO3qZ1dfswpz2/ucilMXGdtM/VzlrxiFPhtkXS6/9SeceZphHZlvEY46J8iIm4vKbT+x7stulHFkptfcXw8p3tbF1k46RXXp+6qxW6ic49G+RWrlHUURcZY6Z+CIe652Qf5cpJ/YvfHyjTHDnnFaba3tXnJafTyc5ZfPH3Ip7snZOX5kpN/c8DHThCM+buzE61EAANGAAAOz1Eno46bC2RsdmfHLSX/B30/EZUV1RdNdkqW5UzlnMH1/rz58zEAN9XFLK41t1VzuqTVdss7oZ/rh9X1PD4jb2XZ7YY7KNX9FLcjGCEt9OvU9ZdLUJKrU2qdu1Zx9WeR34tqtPqqnPfXPUOzKdfaYUeec7v6dCSANr18ZaequzS1TnVDZCxuWUst9M46s628Sesc65Rp0yusVllv1P6lnn4+fRE0AVNZxZz4lZdXGM6XDstslhTj4/tl8zLPWZrurqphVXaopxTbxh58WZQAABKAAAAAAAAAAAAAAAAAAAAAAAAGzQ6v8K88mn1TP3Wav8TbCSlGKh0XP4MQJ5TrSnbry568qtvFO007glFTksN5eP8HLQ65aWDhLbKLeVjKx/YnjqTztve1exj4zXXiWyWrlLWrU7oZXRc+nsd9VxCOor7NYjF9er/4JjWHh9QOU60ntU3E69Pe5R5Q5ffxPABVoAH6008NYYH4AAAAA9VQ7S2MPU8FlJUpV1RXRv2I0JOE4yXVPJWjqKb4qXabJY5pvDMM0WnWnd0dqRM8vbnrqoWaftorElh580ZNBTG657+cYrOPM767VQdXY1PPm10M2jv7C7c/0tYZNIt2/+Vcs4+/Ex6+1XtMW9nhY8v6E/iNMa7IygsKXh9zb2lD+vtljOcZ8SfrdQr7Ft/THp9zPFFos36u2OcfjW9+Nf2aOHUxVTuksy8PsaoyjcnCajJYT5czFoNVGuLqseFnKZrd9FScnYpN+Ty2Rki/Lwv098UYo3Mf8pWor7K+cF0T5FWmqOnpiopOcsZf3JN1jttlN+LKVGqrtqjGyeycfM0yxaa+HL0tscZJ5fp0uqjqKJbopTWcPyZGKuo1VdNLhXLdN+X+SUThi0R5R1c0m/wAf2AA2cgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAVFdbo+EaazSN1ytnPtbIrnlYxHP7c8Es76bW6nSJrT3SgpdUnyf9CEqGn09Fq0j1FU7bddZJOaljZ9WOS8fM906LRqek09lUp2ahSTsU8bcSkk0v6E6vX6uuM4w1E4qbblh9W+rOUdRdGVco2STq/Q8/p555e4FJcP070/4rmqJUpLn0tb24/s2dfwmlWovhVVOuejugt8pZ7T6sc14Px5Ejt7ew7DtJdlu3bM8s+Z0s12qthCFl85Rg04pvxXQClqdFVZrtO9sm9RrLITw/BTS/5Z1hRpoT4bXCnZOd84u1Se7lPBJq4jrKVJV6iyO6W54fj5n5DXaqENsbpqO/elnpLzAoUaLSynpdPOqcrNVBz7ZS/Rza6eOMcz9r0ejl2FDpl2l2mdrt3/AKZYb6eXInQ1uqrolTC+ca5ZzFPz6nhai5SjJWSUoR2ReekfL+4FNcPpXD7XOEYaiqqNvKxt82uqxhcn5nppP/8ALsNZXbk2Wv1c6FTK+brS27c+HkeFqLlqPxHaS7bO7fnnkDvoNPG2y2y2EJU1LM3Obill4XTmULeEUW6idOnzGUJ1yeZZSrklnr5P/JPfFdc5qb1Njklj+h6q4g4V32Sds9VbDsu0lLko8v78gN34Lhi08LZzUIX75QlKUt0Um1HCSw+nPPmYdfVp9PTRCuqXazqhZKbl5rokcadbqaKZU1XTjXLrFPkcZ2Ts275OW2Kis+CXRAeTTRodTfHdXW3HzfI5URU764y6OSTPqboxhThZ5coqMmsvw6EuTqeo7OoiPMvlrqbKJ7bYOL+5zSbeFzZ9DxemPd2ZPMoYxJ9WT+CVwnrcyWdscpBbH1EXxTk16cVw3VuG/sXj9+ZllFxk4yTTXVM+p3y/Ey3dFJRit32648SVx6uMdRCaWJSXMMsHVTkvxtCbVVO6ahXFyl5I728P1VMN86nt8ccypwGuK00rEvqbw2bKN3aOE8v6PrzLOX/wNK5es4XmsR6fKnbT6W7U/wAKDkl4+B+66Ea9ZbCH6VLkfSaaqNeihGt7fpzlBt1HUdqkWj7fNajS3ab+LBxXn4HE+pvhG7h7VnPMM5f7HywT0+fvVmZj0AAOgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAE8PK6os6fjaVajfW3JeKIxUp4O3Up33KrPRBhnjFNfyuPEOIy1mIJba1zx5mbT3z010bK+q/udtdw+zR4k2p1vpJGemmd9sa61mTC9Ipw1X0srjlW3c6XvwStXqZ6u52T/AKLyKC4IsbXqI9pjO0m6nTz01zrsXNf3DHBGCJnte3fQa+ejm+W6EuqN1vG4KtqipqT8ydotFZrJtQ+mK6yZts4L9DdFyslHqgZa9PN/n7Spyc5OUnlt5bKeh4u6KlVdFyiujXUmSTi2msNdShouFWamtWTlsg+nLmw1zRj4fk9PWu4s76nVTHZF9W+pMKGs4XPT1drXNWV+OPAngwxjin4/QAA1AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHTTuMdRW5fpUlk+qtxJQkt/J5TgsnyJso4nqaIbIzTiuiks4Dl6nBOXU1nzCrxaeOGJW47SWOX3J/A5RjrWpdXHCMmo1NupnutlnyXgjlGTjJSi8NdGgnHg44pxzPt9Y8yvScJKMXlPHV4I/H5Rd9cV+pR5nFcX1aht3L98czFOcrJuc25SfVsMun6W2K/KZXeBSi9JOC/UpczZRGUXHMXHbWovPiz5ii+zTz31ScWabuK6q2GxzUU+u1YG1c3R2yXm0T4ly18oz1tsofpcj6Kpu3Q19jjDSXXHLxPlTTpddfpViuf0+T5oN+owTlpERPpfn+Vw6aujGKSfJM+XNOp11+qWLJ/T5LkjME9PhnFWd+5AAHQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAP2EXOaiureEfQVcN02nrirK3bOXIg0z7O2E/S0z6iNkNVTGdfZzT8JeAcPWXvSI4+vtJ4rw6FFaupTUf9UfIxaHSvV6hV5wurf2KvFtRCrRR0ynvm8J/0MHCdTHT6vM+UZLDfkF8Nsk4Nz7+lVaHRJ9l2LfPbu++MkfiOj/B37U8wksxPoOzTu7ZTjjOd3jjHT9iLxrUwv1EY1vKgsZ+4YdJkyWvqZ3GvP8A2/eFcPjqs2252R5JLxKE+HaW6MoRqdU0smfgeqgq5USajLOY58Situn322ShBNc0v8kqdRlzVyzEfp8vdW6bZVy6xeCxouG0w08bdRFzlPpH9yVq7VfqrLV0k+R9Bo9RXq9HFJxckknFvGGQ6urveuOJr+2TW8Npnp5W6eLhKOcx/YiH0urur0einFyTnLOF5tnzQT0lr2pM2/QAA6gAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAOlateeyU/vtyflMO0uhD1NI+oVa08YU0Rivpb5ryx8hz5+ojDrxuZfKyzl7s58cn4XeL0V2aNamMVGSw/wB0zBwjTx1GrxNZjFZx5hametsfcZkr+z5KzZ/XByPrXmNkYJRcX/pxzSx1IfGtNCjURlWsKazhBlg6uMtuOtJ6znl1Pdiuwu0U8eG7JX4HpoOqV8opyzhZ8Cioq5WV2xjKK5ZSCuXra478devb5M9Vqbl+Xu3f7Tpq6lRqrK10i+R9Do9PHTaNOuMXNxy2/ENs2eMVYt/d81YpqX5m7P8AuPJ9PrdPDU6NuUUpbdyfkfMAwZ4zV3AAA3AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAE3vC30w9n8jvC30w9n8kbhOlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlIE3vC30w9n8jvC30w9n8jcGlOMnCSkuqeUX6uJaW+uLum65pYfgfG94W+mHs/kd4W+mHs/kbZZcFcsfJ9NxTiML61RR+jxZj0Oqek1CsSyujX2IveFvph7P5HeFvph7P5G01w1rThEeH2keIaLLu7aSb6xJHEdZ+Mv3JYhFYiiF3hb6Yez+R3hb6Yez+RyUxdLTHPKH0fC+IR026q3PZy8V4FCXEtJp6vyrJWPHKPU+M7wt9MPZ/I7wt9MPZjkjJ0lMluUqt1jutlZLrJ5LOg4pU6FVqHtaWM+aPke8LfTD2fyO8LfTD2fyNtMuCuWvGz67X8UqWndOne5tYz5Ihk3vC30w9n8jvC30w9n8jZiw1xV41UgTe8LfTD2fyO8LfTD2fyNw00pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0pAm94W+mHs/kd4W+mHs/kbg0yAAouAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAP/9k=\n",
      "text/html": [
       "\n",
       "        <iframe\n",
       "            width=\"400\"\n",
       "            height=\"300\"\n",
       "            src=\"https://www.youtube.com/embed/OkkIZNs7Cyc\"\n",
       "            frameborder=\"0\"\n",
       "            allowfullscreen\n",
       "        ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<IPython.lib.display.YouTubeVideo at 0x62ce9bf60>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "YouTubeVideo('OkkIZNs7Cyc')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## CNN en clasificación de imágenes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/html": [
       "\n",
       "        <iframe\n",
       "            width=\"400\"\n",
       "            height=\"300\"\n",
       "            src=\"https://www.youtube.com/embed/l9vg_1YUlzg\"\n",
       "            frameborder=\"0\"\n",
       "            allowfullscreen\n",
       "        ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<IPython.lib.display.YouTubeVideo at 0x62ceb3198>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "YouTubeVideo('l9vg_1YUlzg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Capas en las CNNs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Capa convolucional:\n",
    "**Su función es aumentar la profundidad (espacialidad --> features)**\n",
    "![convolve.png](imgs/convolve.png)\n",
    "- [Ver Ejemplo Auto](https://github.com/udacity/aind2-cnn/blob/master/conv-visualization/conv_visualization.ipynb)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hyperparámetros\n",
    "- Para aumentar la dimensionalidad de una CNN: Aumento la **cantidad de filtros**\n",
    "- Para incrementar el tamaño/complejidad de los patrones detectados: Aumento el **tamaño del filtro**\n",
    "- **Stride**: Paso. Determina el tamaño de las salidas del siguiente nivel\n",
    "- **Padding**: ¿Qué hacemos con los bordes? 'same': mantiene la dimensión de la entrada; 'valid': no padding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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\n",
      "text/html": [
       "\n",
       "        <iframe\n",
       "            width=\"400\"\n",
       "            height=\"300\"\n",
       "            src=\"https://www.youtube.com/embed/Qt5SQNcQfgo\"\n",
       "            frameborder=\"0\"\n",
       "            allowfullscreen\n",
       "        ></iframe>\n",
       "        "
      ],
      "text/plain": [
       "<IPython.lib.display.YouTubeVideo at 0x62ceb3c18>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "YouTubeVideo('Qt5SQNcQfgo')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Como usamos esto en [keras](https://keras.io/layers/convolutional/)?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Convolution2D, Dense"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 28, 28, 1)         28        \n",
      "=================================================================\n",
      "Total params: 28\n",
      "Trainable params: 28\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Convolution2D(filters=1, strides=1, kernel_size=3, padding='same',\n",
    "                        activation='relu',\n",
    "                        input_shape=(28,28,3)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Hacer pruebas y explicar 'Output Shape' y 'Param #':\n",
    "- filters=1, strides=1, kernel_size=1, padding='same'\n",
    "- filters=16, strides=1, kernel_size=1, padding='same'\n",
    "- filters=1, strides=1, kernel_size=4, padding='same'\n",
    "- filters=10, strides=1, kernel_size=4, padding='same'\n",
    "- filters=1, strides=2, kernel_size=5, padding='same'\n",
    "- filters=1, strides=2, kernel_size=5, padding='valid'\n",
    "- filters=16, strides=2, kernel_size=5, padding='same'\n",
    "- filters=16, strides=4, kernel_size=4, padding='same', input_shape=(32,32,1)\n",
    "- filters=16, strides=4, kernel_size=4, padding='same', input_shape=(32,32,3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Convolucionales vs Densas"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Mirar cantidad de **parámetros** y **Output Shape**\n",
    "\n",
    "Las CNN Reducen la cantidad de **parámetros** y la **cantidad de conexiones**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Si hago una analogía con una capa densa, ¿cuántas neuronas tengo a la salida en una CNN?\n",
    "- Multiplico las dimensiones del output-shape\n",
    "\n",
    "Supongamos una capa densa de 1250 neuronas con una entrada de una imagen de 28x28x1. ¿Cuántas conexiones tengo y cuántos parámetros?\n",
    "- 784x1250 = 980.000 conexiones\n",
    "- 784x1250 + 1250 = 981.250 parámetros"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_1 (Dense)              (None, None, 1250)        981250    \n",
      "=================================================================\n",
      "Total params: 981,250\n",
      "Trainable params: 981,250\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(1250, input_shape=(None,784)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para una capa **convolucional** con la misma entrada y la misma cantidad de neuronas a la salida?\n",
    "- Claramente depende del tamaño del kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_2 (Conv2D)            (None, 25, 25, 2)         34        \n",
      "=================================================================\n",
      "Total params: 34\n",
      "Trainable params: 34\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Convolution2D(filters=2, \n",
    "                        strides=1, \n",
    "                        kernel_size=4, \n",
    "                        padding='valid',\n",
    "                        activation='relu',\n",
    "                        input_shape=(28,28,1)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Supongamos un kernel de 4x4 con 2 filtros y padding valid:\n",
    "- (4x4 conexiones por neurona) x (25x25x2 Neuronas) = 20.000 conexiones\n",
    "- (4x4)x2 + 2 = 34 parametros\n",
    "\n",
    "La poca cantidad de parámetros respecto a conexiones se debe obviamente a que comparte parametros"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Se podria entonces hacer que una capa densa tenga la misma salida que una convolucional usando valores de pesos convenientes.\n",
    "\n",
    "La diferencia fundamental es que el aprendizaje es diferente ya que:\n",
    "- Varios de los pesos deberian quedar en cero y no ser modificados por el algoritmo de aprendizaje\n",
    "- Varios de los pesos, los compartidos, deberián moverse iguales"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Entendiendo la salida y los pesos\n",
    "Una vez que aprendió los parámetros"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_simple_model(activation = None, n_filters = 6, input_shape=(28,28,1)):\n",
    "    model = Sequential()\n",
    "    model.add(Convolution2D(filters=n_filters, \n",
    "                            kernel_size=3, \n",
    "                            input_shape=input_shape, \n",
    "                            name = 'Conv1',\n",
    "                            padding= 'valid',\n",
    "                            activation=activation\n",
    "                           ))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "Conv1 (Conv2D)               (None, 26, 26, 4)         220       \n",
      "=================================================================\n",
      "Total params: 220\n",
      "Trainable params: 220\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Tensor de pesos: (3, 3, 6, 4)\n",
      "Biases: (4,)\n"
     ]
    }
   ],
   "source": [
    "canales_entrada = 6\n",
    "model = get_simple_model(n_filters = 4, input_shape=(28,28,canales_entrada))\n",
    "model.summary()\n",
    "print('Tensor de pesos:', model.get_weights()[0].shape)\n",
    "print('Biases:', model.get_weights()[1].shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Modificar los canales de entrada y ver que:\n",
    "- El Output Shape no cambia\n",
    "- La tercer dimensión del tensor de pesos cambia (y por supuesto la cantidad de parámetros)\n",
    "\n",
    "Es decir, la dimensión de los pesos es:  \n",
    "**ancho x alto x canales x filtros**\n",
    "\n",
    "Modificar tambien la cantidad de filtros para verificar como se cambia la ultima dimensión"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Analizando la salida filtrada"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "Conv1 (Conv2D)               (None, 26, 26, 6)         60        \n",
      "=================================================================\n",
      "Total params: 60\n",
      "Trainable params: 60\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Tensor de pesos: (3, 3, 1, 6)\n",
      "Biases: (6,)\n"
     ]
    }
   ],
   "source": [
    "canales_entrada = 1\n",
    "model = get_simple_model(n_filters = 6, input_shape=(28,28,canales_entrada))\n",
    "model.summary()\n",
    "print('Tensor de pesos:', model.get_weights()[0].shape)\n",
    "print('Biases:', model.get_weights()[1].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.04207733 -0.04900354 -0.26046035]\n",
      " [-0.11300875  0.02320629 -0.06815365]\n",
      " [-0.17068511 -0.18162875  0.22570649]]\n",
      "[0. 0. 0. 0. 0. 0.]\n"
     ]
    }
   ],
   "source": [
    "# Matriz de 3x3 inicializada aleatoriamente\n",
    "print(model.get_weights()[0][:,:,0,5])\n",
    "# Biases inicializados en cero\n",
    "print(model.get_weights()[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "from CNN_analysis_helper import plot_filter_coefs, plot_conv_filters, generate_random_image, plot_activations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x576 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_filter_coefs('Conv1', model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights = model.get_layer('Conv1').get_weights()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights[0][:,:,0,0] = np.array([[0, 0, 0], [0 , 1,  0], [0, 0, 0]])\n",
    "weights[0][:,:,0,1] = np.array([[1, 0, -1], [1, 0, -1], [1, 0, -1]])\n",
    "weights[0][:,:,0,2] = np.array([[-1, 0, 1], [-1, 0, 1], [-1, 0, 1]])\n",
    "weights[0][:,:,0,3] = np.array([[1, 0, -1], [1, 0, -1], [1, 0, -1]]).T\n",
    "weights[0][:,:,0,4] = np.array([[-1,-1, 0], [-1 , 0,  1], [0, 1, 1]])\n",
    "weights[0][:,:,0,5] = np.array([[1 , 1,  1], [1, 1, 1], [1, 1, 1]])\n",
    "\n",
    "# Normalizo\n",
    "weights[0] = weights[0]/np.sqrt((weights[0]*weights[0]).sum(axis=0).sum(axis=0))\n",
    "model.get_layer('Conv1').set_weights(weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x576 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_filter_coefs('Conv1', model, normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5, 28, 28, 1)\n",
      "(5, 26, 26, 6)\n"
     ]
    }
   ],
   "source": [
    "images_to_filter = np.load('images_to_filter.npy')\n",
    "print(images_to_filter.shape)\n",
    "activations_conv1_output = model.predict(images_to_filter)\n",
    "print(activations_conv1_output.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sin función de activación"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 30 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_activations(activations_conv1_output)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Con Relu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 30 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_relu = get_simple_model('relu')\n",
    "model_relu.set_weights(weights)\n",
    "activations_output_relu = model_relu.predict(images_to_filter)\n",
    "plot_activations(activations_output_relu)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Con sigmoidea"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 30 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_sigmoid = get_simple_model('sigmoid')\n",
    "model_sigmoid.set_weights(weights)\n",
    "activations_output_sigmoid = model_sigmoid.predict(images_to_filter)\n",
    "plot_activations(activations_output_sigmoid)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sigmoidea normalizando"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(255, 0)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "images_to_filter.max(), images_to_filter.min()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 30 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_sigmoid_2 = get_simple_model('sigmoid')\n",
    "model_sigmoid_2.set_weights(weights)\n",
    "activations_output_sigmoid = model_sigmoid_2.predict(images_to_filter/256)\n",
    "plot_activations(activations_output_sigmoid)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Imagenes grandes con capa densa"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_2 (Dense)              (None, None, 100)         18750100  \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, None, 5)           505       \n",
      "=================================================================\n",
      "Total params: 18,750,605\n",
      "Trainable params: 18,750,605\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(100, input_shape=(None,500*375)))\n",
    "model.add(Dense(5, input_shape=(None,500*375)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Capa Pooling\n",
    "### MaxPooling\n",
    "**Su función es reducir la cantidad de parámetros, bajar la dimensionalidad**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Max Pooling\n",
    "![Max-polling.png](imgs/Max-polling.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Hyperparámetros\n",
    "MaxPooling\n",
    "- Pool size\n",
    "- Strides: En este caso está en None por default lo que significa que strides=pool_size"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### En [keras](https://keras.io/layers/pooling/)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import MaxPooling2D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "max_pooling2d_1 (MaxPooling2 (None, 16, 16, 1)         0         \n",
      "=================================================================\n",
      "Total params: 0\n",
      "Trainable params: 0\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(MaxPooling2D(pool_size=(2,2), strides=None, padding='same', input_shape=(32,32,1)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Hacer pruebas y explicar 'Output Shape' y 'Param #':\n",
    "- pool_size=(2,2), strides=None, padding='same'\n",
    "- pool_size=(4,2), strides=None, padding='same'\n",
    "- pool_size=(3,3), strides=None, padding='same'\n",
    "- pool_size=(3,3), strides=None, padding='valid'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_simple_model_pool(activation = None, n_filters = 6, input_shape=(28,28,1)):\n",
    "    model = Sequential()\n",
    "    model.add(Convolution2D(filters=n_filters, \n",
    "                            kernel_size=3, \n",
    "                            input_shape=input_shape, \n",
    "                            name = 'Conv1',\n",
    "                            padding= 'valid',\n",
    "                            activation=activation\n",
    "                           ))\n",
    "    model.add(MaxPooling2D(3))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_pool = get_simple_model_pool(activation='relu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_pool.get_layer('Conv1').set_weights(weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5, 8, 8, 6)\n"
     ]
    }
   ],
   "source": [
    "activations_pool_output = model_pool.predict(images_to_filter)\n",
    "print(activations_pool_output.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 30 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_activations(activations_pool_output)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Global Average Pooling  \n",
    "- Un resultado(Promedio) por cada canal\n",
    "- Utilizada para reducir la dimensionalidad en las ultimas capas\n",
    "![GAP.png](imgs/GAP.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### En [keras](https://keras.io/layers/pooling/)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import GlobalAveragePooling2D, Flatten"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "global_average_pooling2d_1 ( (None, 20)                0         \n",
      "=================================================================\n",
      "Total params: 0\n",
      "Trainable params: 0\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(GlobalAveragePooling2D(input_shape=(32,32,20)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### AveragePooling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import AveragePooling2D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "average_pooling2d_1 (Average (None, 8, 8, 20)          0         \n",
      "=================================================================\n",
      "Total params: 0\n",
      "Trainable params: 0\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(AveragePooling2D(4, 4, input_shape=(32,32,20)))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preguntas:\n",
    "\n",
    "- Como cierro la red? Que utlilzar? GAP o DENSA?\n",
    "- Que implica hacer mas grande o mas pequeño el filtro?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Armamos la red AlexNet:"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### AlexNet [(paper)](http://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf)\n",
    "- 2012 \n",
    "- Error de top-5 de 15.3%, más de 10.8% arriba del segundo. \n",
    "- ImageNet: En 1.2 millones de imágenes de alta resolución, 1000 clases\n",
    "- Diseñada por: SuperVision group, consisting of Alex Krizhevsky, Geoffrey Hinton, and Ilya Sutskever.\n",
    "\n",
    "**Lo novedoso**: Dropout, RELUs\n",
    "(Dropout a la salida de las capas de 2048)\n",
    "\n",
    "**Arquitectura:**\n",
    "- Rectangulares con center-crop (Lado mas corto 256)\n",
    "- 60 millones de parámetros\n",
    "- 650 mil neuronas\n",
    "- 5 capas convolucionales\n",
    "- Algunos seguidos de capas pooling\n",
    "- 3 capas densas\n",
    "- SoftMax de 1000 categorias (1000 neuronas a la salida)\n",
    "- RELUs en vez de Sigmoideas\n",
    "- Dropout para evitar regularización\n",
    "\n",
    "**Entrenamiento:**\n",
    "- CUDA para soporte de GPU\n",
    "- Data augmentation\n",
    "- SGD\n",
    "- Batch size = 128\n",
    "- Momentum 0.9\n",
    "- Learning rate = 0.01\n",
    "- weight decay of 0.0005 (equivalente a L2 regularization)\n",
    "- 5 a 6 días de entrenamiento en 2 [GTX 580](https://www.geforce.com/hardware/desktop-gpus/geforce-gtx-580) 3GB GPUs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![alexnet2.png](imgs/alexnet2.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Armarla con Keras:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "from solutions import get_alexnet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_3 (Conv2D)            (None, 55, 55, 96)        34944     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 27, 27, 96)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 27, 27, 256)       614656    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 13, 13, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 13, 13, 384)       885120    \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 13, 13, 384)       1327488   \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 6, 6, 256)         884992    \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 9216)              0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 4096)              37752832  \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 4096)              0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 4096)              16781312  \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 4096)              0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 1000)              4097000   \n",
      "=================================================================\n",
      "Total params: 62,378,344\n",
      "Trainable params: 62,378,344\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "alexnet_model = Sequential()\n",
    "alexnet_model.add(Convolution2D(filters=96, kernel_size=11, strides=4,\n",
    "                                input_shape=(227,227,3), \n",
    "                                padding='valid', activation='relu'))\n",
    "alexnet_model.add(MaxPooling2D(pool_size = 3, strides=2))\n",
    "alexnet_model.add(Convolution2D(filters=256, kernel_size=5, padding='same', activation='relu'))\n",
    "alexnet_model.add(MaxPooling2D(pool_size = 3, strides=2))\n",
    "alexnet_model.add(Convolution2D(filters=384, kernel_size=3, padding='same', activation='relu'))\n",
    "alexnet_model.add(Convolution2D(filters=384, kernel_size=3, padding='same', activation='relu'))\n",
    "alexnet_model.add(Convolution2D(filters=256, strides=2, kernel_size=3, padding='valid', activation='relu'))\n",
    "alexnet_model.add(Flatten())\n",
    "alexnet_model.add(Dense(4096, activation='relu'))\n",
    "alexnet_model.add(Dropout(0.5))\n",
    "alexnet_model.add(Dense(4096, activation='relu'))\n",
    "alexnet_model.add(Dropout(0.5))\n",
    "alexnet_model.add(Dense(1000, activation='softmax'))\n",
    "alexnet_model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Reemplazo de FC por Convolucionales e implementación de sliding window eficiente"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Supongamos que tengo una entrada de 5x5. Si hago flatten para poner una FC, me queda una entrada de 25x1.  \n",
    "La cantidad de parámetros a aprender suponiendo que tengo una salida con 4 neuronas es:  \n",
    "25*4 + 4 = 104"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_7 (Dense)              (None, 4)                 104       \n",
      "=================================================================\n",
      "Total params: 104\n",
      "Trainable params: 104\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.layers import Dense\n",
    "model_dense = Sequential()\n",
    "model_dense.add(Dense(4, input_shape=(25,)))\n",
    "model_dense.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Si utilizo una convolucional con la misma entrada que salida, para tener la misma cantidad de parámetros, tengo que poner el tamaño del filtro igual a la entrada: 5x5  \n",
    "Y la cantidad de filtros igual a la salida"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_8 (Conv2D)            (None, 1, 1, 4)           104       \n",
      "=================================================================\n",
      "Total params: 104\n",
      "Trainable params: 104\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model_slide = Sequential()\n",
    "model_slide.add(Convolution2D(filters= 4, kernel_size=5, input_shape=(5,5,1)))\n",
    "model_slide.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Que pasa si aumento la entrada:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_9 (Conv2D)            (None, 2, 2, 4)           104       \n",
      "=================================================================\n",
      "Total params: 104\n",
      "Trainable params: 104\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "width = 6\n",
    "height = width\n",
    "model_slide = Sequential()\n",
    "model_slide.add(Convolution2D(filters= 4, kernel_size=5, input_shape=(height,width,1)))\n",
    "model_slide.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![](imgs/sliding-wind-implem.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<pre>Layer         kernel  stride  output shape\n",
    "---------------------------------------------\n",
    "Input                          (416, 416, 3)\n",
    "Convolution    3×3      1      (416, 416, 16)\n",
    "MaxPooling     2×2      2      (208, 208, 16)\n",
    "Convolution    3×3      1      (208, 208, 32)\n",
    "MaxPooling     2×2      2      (104, 104, 32)\n",
    "Convolution    3×3      1      (104, 104, 64)\n",
    "MaxPooling     2×2      2      (52, 52, 64)\n",
    "Convolution    3×3      1      (52, 52, 128)\n",
    "MaxPooling     2×2      2      (26, 26, 128)\n",
    "Convolution    3×3      1      (26, 26, 256)\n",
    "MaxPooling     2×2      2      (13, 13, 256)\n",
    "Convolution    3×3      1      (13, 13, 512)\n",
    "MaxPooling     2×2      1      (13, 13, 512)\n",
    "Convolution    3×3      1      (13, 13, 1024)\n",
    "Convolution    3×3      1      (13, 13, 1024)\n",
    "Convolution    1×1      1      (13, 13, 125)\n",
    "</pre>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  },
  "toc": {
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": true,
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": "block",
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}