caffe_cifar10_adversarial.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CIFAR 10 - adversarial networks\n",
    "\n",
    "This notebook contains analysis of several 5 layer networks trained in caffe on the CIFAR 10 data set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# imports and basic notebook setup\n",
    "from cStringIO import StringIO\n",
    "import numpy as np\n",
    "import scipy.ndimage as nd\n",
    "import PIL.Image\n",
    "from IPython.display import clear_output, Image, display\n",
    "from google.protobuf import text_format\n",
    "import matplotlib.pyplot as plt\n",
    "from cs231n.data_utils import load_CIFAR10\n",
    "import caffe\n",
    "\n",
    "%matplotlib inline\n",
    "plt.rcParams['figure.figsize'] = (10.0, 8.0) # set default size of plots\n",
    "plt.rcParams['image.interpolation'] = 'nearest'\n",
    "plt.rcParams['image.cmap'] = 'gray'\n",
    "\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "\n",
    "# If your GPU supports CUDA and Caffe was built with CUDA support,\n",
    "# uncomment the following to run Caffe operations on the GPU.\n",
    "caffe.set_mode_gpu()\n",
    "#caffe.set_device(0) # select GPU device if multiple devices exist\n",
    "\n",
    "def showarray(a, fmt='jpeg'):\n",
    "    a = np.uint8(np.clip(a, 0, 255))\n",
    "    f = StringIO()\n",
    "    PIL.Image.fromarray(a).save(f, fmt)\n",
    "    display(Image(data=f.getvalue()))\n",
    "    \n",
    "\n",
    "def rel_error(x, y):\n",
    "  \"\"\" returns relative error \"\"\"\n",
    "  return np.max(np.abs(x - y) / (np.maximum(1e-8, np.abs(x) + np.abs(y))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Load the model\n",
    "\n",
    "model_path = '/Users/amaharaj/Documents/Courses/neural_networks/project/assignment2/' # substitute your path here\n",
    "net_fn   = model_path + 'cifar10_deploy.prototxt'\n",
    "param_fn = model_path + 'cifar10_disc4.caffemodel.h5'\n",
    "\n",
    "# Patching model to be able to compute gradients.\n",
    "# Note that you can also manually add \"force_backward: true\" line to \"deploy.prototxt\".\n",
    "\n",
    "net = caffe.Classifier(net_fn, param_fn,\n",
    "                       channel_swap = (2,1,0)) # the reference model has channels in BGR order instead of RGB\n",
    "\n",
    "# a couple of utility functions for converting to and from Caffe's input image layout\n",
    "def preprocess(net, img):\n",
    "    return np.float32(np.rollaxis(img, 2)[::-1]) - net.transformer.mean['data']\n",
    "def deprocess(net, img):\n",
    "    return np.dstack((img + net.transformer.mean['data'])[::-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_test:  (10000, 3, 32, 32)\n",
      "X_train:  (50000, 3, 32, 32)\n",
      "y_train:  (50000,)\n",
      "y_test:  (10000,)\n"
     ]
    }
   ],
   "source": [
    "# Load the (preprocessed) CIFAR10 data.\n",
    "rawdata={}\n",
    "rawdata['X_train'],rawdata['y_train'],rawdata['X_test'],rawdata['y_test'] = load_CIFAR10('cs231n/datasets/cifar-10-batches-py')\n",
    "\n",
    "\n",
    "rawdata['X_train']=rawdata['X_train'].transpose(0,3,1,2) \n",
    "rawdata['X_test']=rawdata['X_test'].transpose(0,3,1,2) \n",
    "\n",
    "for k, v in rawdata.iteritems():\n",
    "    print '%s: ' % k, v.shape\n",
    "\n",
    "#the number of discretizations\n",
    "pts = 4\n",
    "rawdata['X_train_disc'] = (rawdata['X_train']//(256/pts))*(256/pts) + 0.5*(256/pts)\n",
    "rawdata['X_test_disc'] = (rawdata['X_test']//(256/pts))*(256/pts) + 0.5*(256/pts)\n",
    "rawdata['X_train_disc'][rawdata['X_train_disc'] > 255.0] = 255.0\n",
    "rawdata['X_test_disc'][rawdata['X_test_disc'] > 255.0] = 255.0\n",
    "\n",
    "# Subtract means from the data\n",
    "meandisc_image = np.mean(rawdata['X_train_disc'], axis=0)\n",
    "mean_image = np.mean(rawdata['X_train'], axis=0)\n",
    "rawdata['X_train']-= mean_image\n",
    "rawdata['X_test']-= mean_image\n",
    "rawdata['X_train_disc']-= meandisc_image\n",
    "rawdata['X_test_disc']-= meandisc_image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x1cf7d2c10>"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVMAAAFRCAYAAADJg05/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAGKpJREFUeJzt3X9oVff9x/FXrppWo7b+IUNauYnWsdkfli0d+ady/ojD\nUNlm2a8O6zWGTWUD64/VaKKJGrVIGQNRaNcRMgXDRv9QBqWtW88KI9RSUKbWDrb0FmImdXTVJKNJ\nmvP9o18ye++5N5+c+74/jj4ff/V+fJ/P593j9cVJ7vncUxUEQSAAQEES5W4AAO4EhCkAGCBMAcAA\nYQoABghTADBAmAKAgZlRDgqCQJ2dnfrggw9UXV2tQ4cOafHixTnr0+l01tiiRYs0ODgYZfmKEK/+\n/ayRRYu+o8HBs6VvxUix+/cd67yI89+Z598rRyuRFPLvN5lMho5HujI9d+6cRkdH1dvbqx07dujI\nkSPTnqO6ujrK0hUj/v0vKHcLBaH/8op///b/fiOF6Xvvvacnn3xSkrRixQpdunTJtCkAiJtIYTo0\nNKR58+ZNvp45c6YmJibMmgKAuIkUpnPnztXw8PDk64mJCSUSfJYF4O5VFWVv/htvvKG33npLR44c\n0YULF3TixAm9/PLLOetHR0dj/ztGAEin0zk/gIr0af6qVav017/+VT/+8Y8lacoPoMI+NUsmk6Gf\n8sdFvPr3s0aSyZTS6Z7St2Kk2P37jnVexPnvzPPvlaOVSIrx7zdSmFZVVWn//v2mjQBAnPGLTgAw\nEOnKFHHjOY77Re3iTuQ71HhF7gGVgStTADBAmAKAAcIUAAwQpgBggDAFAAOEKQAYIEwBwABhCgAG\nuGnfiR8ylsoa98PKIvI8z26yu4RfhjW9MqxpyyvTsXcerkwBwABhCgAGCFMAMECYAoABwhQADBCm\nAGCAMAUAA4QpABggTAHAADugSsx33CblG+7n8UJ2qqRSqaxeKnXTlR8ylr3/rMAJSzxVKmW7Yy46\nP9JRYe8fV27vM6eiisKVKQAYIEwBwABhCgAGCFMAMECYAoABwhQADBCmAGCAMAUAA4QpABioCoIg\nKPYi6XQ6ayyZTIaOVyY/aySZTCmd7pmiymmq8DLXnVIRd6H09/errq7uS2Muz50qx7Opwv4fu7u7\n1dzcbDZfqYWd/zgppH/L95DrXJl1heRPMpkMHefKFAAMEKYAYIAwBQADhCkAGCBMAcAAYQoABghT\nADBAmAKAAR5b4sRzGs9VdTvXx5FY3thsuQHA+YZ315upHWpyrVkJN99j+iz/3lzn6uzs/NLr8Mf2\neAX1EjlMn376ac2dO1eS9OCDD+rw4cMFNQIAcRYpTEdHRyVJv/vd70ybAYC4ivQ706tXr2pkZEQt\nLS3asGGDLl68aN0XAMRKpCvTe++9Vy0tLfrBD36gDz/8UD/96U/1+uuvK5Hg8ywAd6dIYVpbWzv5\nzSm1tbW6//779fHHH+srX/mKaXMAEBeRvoLv9OnT+vvf/66Ojg5dv35dzc3N+uMf/5jzynR0dFTV\n1dUFNwsA5ZROp3N+BV+kMB0bG9Pu3bt17do1JRIJ7dy5U48//njeBjLF6/tMs0Xt3/p2nqi3MxX9\n+zSLfGvU3fx9oJUgbv2H3RrV0/Pl7yN2vTUqV5hG+jF/1qxZevHFF6McCgB3JD4xAgAD7IAqMddN\nFp3OO41cSsKLMn+scdqd5VDyxZqOdQXsOsnqnx1RyCHsx/zMMdf3Ynd3d+g4V6YAYIAwBQADhCkA\nGCBMAcAAYQoABghTADBAmAKAAcIUAAzE8KZ9v9wN/L+UovTifIRjodON6jluRs4+MrxuqqOiV7mt\nmOtmam7ah6VC3z9cmQKAAcIUAAwQpgBggDAFAAOEKQAYIEwBwABhCgAGCFMAMECYAoCBEu2A8kPG\nou0gij3fscxyN0+uuSpgx1DUDsLePa6PnYi6prkCHtlSchXwXql0XJkCgAHCFAAMEKYAYIAwBQAD\nhCkAGCBMAcAAYQoABghTADBAmAKAgRg+A6py+S417CSZNs9x3HfdURT3v4O493+H4soUAAwQpgBg\ngDAFAAOEKQAYIEwBwABhCgAGCFMAMECYAoCBu/qmfd+xzjOckJv2i8dzLnSodPx7Kujvk/fCHcXp\nyvTixYt69tlnJUkfffSRfvKTn2jdunXav39/UZsDgLiYMkxfeeUVtbe3a2xsTJJ05MgRbd++XadO\nndLExITOnTtX9CYBoNJNGabJZFLHjx+ffH358mXV19dLklauXKm+vr7idQcAMTFlmK5atUozZsyY\nfB0EweR/19TU6NatW8XpDABiZNqf5icS/ztkeHhY8+fPN20IAOKoKrj9UjOHgYEB7dixQ729vdqy\nZYs2btyoJ554Qh0dHWpoaFBTU1Pe40dHP1F19QKzpgGgHOrq6tTf3x/6Z9O+NWrXrl3au3evxsbG\ntHTpUq1evXrKYwYHz2aNJZMppdM9013elO9Y54WMhfXvcqdLZ2en46rF1d/fr7q6uuIt4Pjdop5D\nXVhFKpVST0+094/vVORUFfnWqKKf/yKj/2xOYfrAAw+ot7dXklRbW6uTJ0+aNgEAcccOKAAwUFE7\noHyHGq9C5kqFjJvubvLMC0MOzTi2Qnfk+CFjYeffsn+XXz+48nO9GzOXyFGGeODKFAAMEKYAYIAw\nBQADhCkAGCBMAcAAYQoABghTADBAmAKAAcIUAAxU1A4oz6HGL3IPBfFcShyKjPmuhYa7fix3EIXu\nbEqlssZDqkJ5Tku6zeb2RS3hNZ1e55fXNHx3O89lt+RdjytTADBAmAKAAcIUAAwQpgBggDAFAAOE\nKQAYIEwBwABhCgAGKuqm/VLzHes813HTR124FjpXmijHpgNXXhnW9B1qPMe5XDc6uNyQ7/mOqzqW\nha6ZeWxIyd2EK1MAMECYAoABwhQADBCmAGCAMAUAA4QpABggTAHAAGEKAAYIUwAwUFE7oPxyN1BR\n/JKv6JkVTUOJd3DZ8x1qPLepXMscdje5boByfZ+FrZm5G855zdj/nYfjyhQADBCmAGCAMAUAA4Qp\nABggTAHAAGEKAAYIUwAwQJgCgIGS3LTvh4ylcozf6fwy3LDsTXMcmsbJcS6047CkQ8kXHO+0Dy3L\neMyK5/je9l0ezxLDG/udrkwvXryoZ599VpL0/vvva+XKlVq/fr3Wr1+v1157ragNAkAcTHll+sor\nr+jMmTOqqamRJF26dEkbN27Uhg0bit0bAMTGlFemyWRSx48fn3x9+fJl+b6vdevWqa2tTSMjI0Vt\nEADiYMowXbVqlWbMmDH5esWKFXr++ed16tQpLV68WMeOHStqgwAQB9P+NL+xsVHLly+X9EXQXr16\n1bwpAIibqiAIgqmKBgYGtGPHDvX29uqHP/yh9u7dq0cffVSnTp3Sv/71L+3cuTPv8Z+MfqIF1QvM\nmgaAcqirq1N/f3/on0371qjOzk4dPHhQs2bN0sKFC3XgwIEpjzk7eDZrLJVMqSfdM93ly8ILGUsm\nU0pn9O87zOV3ulQVX3d3t5qbm6d/oMttLdMR8RaYyP278uwKwypSqZR6enqmLiw2P1pZdyql5sz+\nXW+Ncipymyuq/v5+1dXVmc7pFKYPPPCAent7JUnLly/X6dOnTZsAgLhjBxQAGCjJDihvmuP5+JG7\niBffsc4rYg93LM+lxKGo0PW8vC9zMv0J2PHXNmG7mzKPzK7IMZdDjdMuqekowY4qrkwBwABhCgAG\nCFMAMECYAoABwhQADBCmAGCAMAUAA4QpABggTAHAQEl2QLnyy91AoXyXEoci2e7AybVi5rjdisY8\n1/FchRGq3KYqaL3Mcd9uydjzjOcL3VGVNeYXtAZXpgBggDAFAAOEKQAYIEwBwABhCgAGCFMAMECY\nAoABwhQADFTUTful5hnP53JDvunjMMrAc6zznSd0nTH04DyvnA+LWlLZXM+r6xNFM+pSqVTWWCXz\nnMbCqtxxZQoABghTADBAmAKAAcIUAAwQpgBggDAFAAOEKQAYIEwBwABhCgAGSrIDyg8ZS+UYr0R+\nyFho/2GFxea5lIQXeRm7ZBymKgvPdTxXYbQyJ37E4wp6/3tRD0QxcWUKAAYIUwAwQJgCgAHCFAAM\nEKYAYIAwBQADhCkAGCBMAcAAYQoABvLugBofH9eePXs0MDCgsbExbd68WQ899JBaW1uVSCS0bNky\ndXR0lKrXafEsJ/NDxsK2sDgt6lSUY9Fo8+WqyBqfeirntlymcpZrMi/vy4rhOY77pu8NlxpYyhum\nZ8+e1YIFC3T06FHdvHlT3/3ud/W1r31N27dvV319vTo6OnTu3Dk1NjaWql8AqEh5f8xvamrS1q1b\nJUmff/65ZsyYoStXrqi+vl6StHLlSvX19RW/SwCocHnDdPbs2ZozZ46Ghoa0detWbdu2TUEQTP55\nTU2Nbt26VfQmAaDSTfkB1ODgoFKplNauXaunnnpKicT/DhkeHtb8+fOL2iAAxEFVcPulZoYbN25o\n/fr12rdvnxoaGiRJW7Zs0caNG/XEE0+oo6NDDQ0NampqyrvIJ6OfaEH1AtvOAaDEmpub1d3dHfpn\neT+Aeumll3Tz5k2dOHFCx48fV1VVldra2tTV1aWxsTEtXbpUq1evnrKBs4Nns8ZSyZR60j2O/wvT\n5xnO5fvZY6lUSj09X+7fd/oE1XNd1bFu6vnCKsL6t/w035SXPRT2/gkpKzrfocYLGUsmU0qnM98/\nYZVRV7Wdys/4R9Dd3a3m5ma7PkqsGP3nDdO2tja1tbVljZ88edK0CQCIO27aBwADJXlsiTfN8Yrj\nuY17fq5Cy0W/zC9kKrclCj+mwOl8x7lc6yx5BVV5eV7lFvZrJ5QfV6YAYIAwBQADhCkAGCBMAcAA\nYQoABghTADBAmAKAAcIUAAwQpgBgoCQ7oOLOcx3PVRiB7S4Xz2k8V1U0vulslcsrdwOh3L50R/Iq\ntP+y8Ao7nCtTADBAmAKAAcIUAAwQpgBggDAFAAOEKQAYIEwBwABhCgAGuGm/UnmOZb5LVVhRKmvc\nZSrPZbm7iOvN8ZlSSkU+1pJzB5439VglP0/FcxwrAFemAGCAMAUAA4QpABggTAHAAGEKAAYIUwAw\nQJgCgAHCFAAMEKYAYIAdUHcDP2QsewMU25vuJL5toReyAypzxM8uKWRJN65rlgBXpgBggDAFAAOE\nKQAYIEwBwABhCgAGCFMAMECYAoABwhQADBCmAGAg7w6o8fFx7dmzRwMDAxobG9PmzZu1aNEibdq0\nSbW1tZKkZ555Rk1NTaXo9a7imReGHFrAsVb8cjdwp/LKsaTbor7Lg8scSsy5rpkKH84bpmfPntWC\nBQt09OhRffrpp/re976nn//859q4caM2bNgwrT4B4E6WN0ybmpq0evVqSdLExIRmzpypy5cv65//\n/KfOnTunZDKptrY2zZkzpyTNAkClyvs709mzZ2vOnDkaGhrS1q1b9dxzz+mxxx7Trl27dOrUKS1e\nvFjHjh0rVa8AULGqgiAI8hUMDg7qF7/4hdatW6e1a9fq1q1bmjdvniTpH//4h7q6utTd3Z13kdHR\nT1RdvcCuawAog+bm5px5l/fH/Bs3bqilpUX79u1TQ0ODJKmlpUV79+7Vo48+qr6+Pj388MNTNjA4\neDZrLJlMKZ3ucem/ItG/DT/icalkSj0V0H9UBfXvu5Q4FE1D5odLqVRKPT3R+nfqzaFEUuQP2rpT\n3WruaY62Zg55w/Sll17SzZs3deLECR0/flxVVVXavXu3Dh8+rFmzZmnhwoU6cOBAYR0AwB0gb5i2\ntbWpra0ta/z06dNFawgA4oib9gHAQIkeW+I5jvsFzBWFy3ooJs+hxnedzLmwxLxyN1A4P+N1IU+9\ncbm53+nG/grDlSkAGCBMAcAAYQoABghTADBAmAKAAcIUAAwQpgBggDAFAAOEKQAYKNEOKD9kLGwP\nRbHXvDv5IWPFPvuuvHI3AEd+xusC9kA5lHm+41wFyNyJVeiuK65MAcAAYQoABghTADBAmAKAAcIU\nAAwQpgBggDAFAAOEKQAYIEwBwECJdkBZ8svdQEH8Slm0QrZAmW50sZzLkh8yVshDlMrCcxxz4Efv\nophcnk2VD1emAGCAMAUAA4QpABggTAHAAGEKAAYIUwAwQJgCgAHCFAAMlOimfW+a4/n4kbuwnCny\nPe+FLBonnmOdH3H+Ctl0YMqv4DW9IvZQCK+AQs/5YCdcmQKAAcIUAAwQpgBggDAFAAOEKQAYIEwB\nwABhCgAGCFMAMECYAoCBKXdATUxMqL29Xf39/UokEtq/f7+qq6vV2tqqRCKhZcuWqaOjw6gd36Ci\nwnnlbuA2XsZr37eb2/l5JFHXrJQtUF65GwjnF3vC7PPvumShjweJImzFsLFCTBmmf/7zn1VVVaXT\np0/r/Pnz+tWvfqUgCLR9+3bV19ero6ND586dU2Njo3FrABAfU/6Y39jYqIMHD0qSrl27pvvuu09X\nrlxRfX29JGnlypXq6+srbpcAUOGcfmeaSCTU2tqqrq4urVmzRkEQTP5ZTU2Nbt26VbQGASAOqoLb\nk3EK//73v/X9739fIyMjeueddyRJf/rTn9TX16f29vacx42Ojqq6urrwbgGgjNLptJLJZOifTfk7\n0zNnzuj69ev62c9+pnvuuUeJREKPPPKIzp8/r29961t6++231dDQkHeOwcHBrLFkMql0Op0x6k/V\nTkV85CBJqWRKPemecrcRWWj/lh9AFfk7+FKpbvX0NEc61pYX6ahUKqWenuK9f3zTv0tl/W92p7rV\nHPH8m34A5ThV5prh+VOYKcP029/+tnbv3q1169ZpfHxc7e3tWrJkidrb2zU2NqalS5dq9erVpk0B\nQNxMGaazZ8/Wr3/966zxkydPFqUhAIgjbtoHAAMlemyJHzIW/aZfO4WsWCk3jUcV0r9nOb8/ZcV0\nykx5lpP5DiWmC7opy5JlWLSCcGUKAAYIUwAwQJgCgAHCFAAMEKYAYIAwBQADhCkAGCBMAcAAYQoA\nBqb1FXwAgHBcmQKAAcIUAAwQpgBggDAFAAOEKQAYIEwBwECJvhz6f4IgUGdnpz744ANVV1fr0KFD\nWrx4canbKMjTTz+tuXPnSpIefPBBHT58uMwdubl48aJefPFFnTx5Uh999JFaW1uVSCS0bNkydXR0\nlLu9Kd3e//vvv69NmzaptrZWkvTMM8+oqampvA3mMD4+rj179mhgYEBjY2PavHmzHnroodic/7D+\nFy1aFJvzPzExofb2dvX39yuRSGj//v2qrq62P/9Bib3xxhtBa2trEARBcOHChWDLli2lbqEgn332\nWbB27dpytzFtv/nNb4I1a9YEP/rRj4IgCILNmzcH7777bhAEQbBv377gzTffLGd7U8rs//e//33Q\n3d1d3qYcvfrqq8Hhw4eDIAiCTz/9NPA8L1bn//b+//Of/wSe5wV/+MMfYnP+33zzzWDPnj1BEATB\nO++8E2zZsqUo57/kP+a/9957evLJJyVJK1as0KVLl0rdQkGuXr2qkZERtbS0aMOGDbp48WK5W3KS\nTCZ1/PjxydeXL19WfX29JGnlypXq6+srV2tOwvr3fV/r1q1TW1ubRkZGythdfk1NTdq6dask6fPP\nP9eMGTN05cqV2Jz/2/ufmJjQzJkzdfnyZb311luxOP+NjY06ePCgJOnatWu67777inL+Sx6mQ0ND\nmjdv3uTrmTNnamJiotRtRHbvvfeqpaVFv/3tb9XZ2amdO3fGov9Vq1ZpxowZk6+D2za+1dTU6Nat\nW+Voy1lm/ytWrNDzzz+vU6dOafHixTp27FgZu8tv9uzZmjNnjoaGhrR161Zt27YtVuc/s//nnntO\njz32mHbt2hWL8y9JiURCra2t6urq0po1a4py/ksepnPnztXw8PDk64mJCSUS8fkcrLa2Vt/5zncm\n//v+++/Xxx9/XOaupu/2cz48PKz58+eXsZvpa2xs1PLlyyV9EbRXr14tc0f5DQ4OKpVKae3atXrq\nqadid/4z+4/b+ZekF154Qa+//rra29v12WefTY5bnf+Sp9g3vvEN/eUvf5EkXbhwQV/96ldL3UJB\nXn31Vb3wwguSpOvXr2t4eFgLFy4sc1fTt3z5cr377ruSpLffflvf/OY3y9zR9LS0tOhvf/ubJKmv\nr08PP/xwmTvK7caNG2ppadEvf/lLrV27VpL09a9/PTbnP6z/OJ3/M2fO6OWXX5Yk3XPPPUokEnrk\nkUd0/vx5SXbnv+RfdBLc9mm+JB05ckR1dXWlbKEgY2Nj2r17t65du6ZEIqGdO3fq8ccfL3dbTgYG\nBrRjxw719vbqww8/1N69ezU2NqalS5eqq6tLVVVV5W4xr9v7v3Llig4ePKhZs2Zp4cKFOnDggGpq\nasrdYqhDhw7ptdde05IlSxQEgaqqqtTW1qaurq5YnP+w/rdt26ajR4/G4vz/97//1e7du3Xjxg2N\nj49r06ZNWrJkidrb203PP98aBQAG4vPLSgCoYIQpABggTAHAAGEKAAYIUwAwQJgCgAHCFAAMEKYA\nYOD/AKqumqcJWOmZAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1cf3dec90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Show an example of the mean subtracted data\n",
    "plt.imshow((rawdata['X_train_disc'][1113] + meandisc_image).transpose(1,2,0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "distortion: 0.000000, accuracy: 0.751700\n",
      "distortion: 2.560000, accuracy: 0.751700\n",
      "distortion: 5.120000, accuracy: 0.751700\n",
      "distortion: 7.680000, accuracy: 0.751700\n",
      "distortion: 10.240000, accuracy: 0.751700\n",
      "distortion: 12.800000, accuracy: 0.751700\n",
      "distortion: 15.360000, accuracy: 0.751700\n",
      "distortion: 17.920000, accuracy: 0.751700\n",
      "distortion: 20.480000, accuracy: 0.751700\n",
      "distortion: 23.040000, accuracy: 0.751700\n",
      "distortion: 25.600000, accuracy: 0.751700\n",
      "distortion: 28.160000, accuracy: 0.751700\n",
      "distortion: 30.720000, accuracy: 0.751700\n",
      "distortion: 33.280000, accuracy: 0.069100\n",
      "distortion: 35.840000, accuracy: 0.149900\n",
      "distortion: 38.400000, accuracy: 0.123600\n",
      "distortion: 40.960000, accuracy: 0.166600\n",
      "distortion: 43.520000, accuracy: 0.139900\n",
      "distortion: 46.080000, accuracy: 0.150000\n",
      "distortion: 48.640000, accuracy: 0.137400\n",
      "distortion: 51.200000, accuracy: 0.153000\n",
      "distortion: 53.760000, accuracy: 0.138200\n",
      "distortion: 56.320000, accuracy: 0.144800\n",
      "distortion: 58.880000, accuracy: 0.147800\n",
      "distortion: 61.440000, accuracy: 0.142700\n",
      "distortion: 64.000000, accuracy: 0.143800\n",
      "distortion: 66.560000, accuracy: 0.143400\n",
      "distortion: 69.120000, accuracy: 0.144300\n",
      "distortion: 71.680000, accuracy: 0.144600\n",
      "distortion: 74.240000, accuracy: 0.147700\n",
      "distortion: 76.800000, accuracy: 0.142200\n",
      "distortion: 79.360000, accuracy: 0.143200\n",
      "distortion: 81.920000, accuracy: 0.146100\n",
      "distortion: 84.480000, accuracy: 0.143100\n",
      "distortion: 87.040000, accuracy: 0.145000\n",
      "distortion: 89.600000, accuracy: 0.141400\n",
      "distortion: 92.160000, accuracy: 0.141100\n",
      "distortion: 94.720000, accuracy: 0.143700\n",
      "distortion: 97.280000, accuracy: 0.116500\n",
      "distortion: 99.840000, accuracy: 0.111100\n"
     ]
    }
   ],
   "source": [
    "#Now perform tests as we increase the strength of distortion\n",
    "\n",
    "y_test = rawdata['y_test']\n",
    "distortion = np.zeros(rawdata['X_test'].shape)\n",
    "epsilonvec=np.zeros(40)\n",
    "\n",
    "for i in range(40):\n",
    "    epsilon = 256*0.01*i\n",
    "    y_pred = np.zeros(10000)\n",
    "    \n",
    "    #Add the distortion\n",
    "    distorted_data = rawdata['X_test_disc'] - epsilon*distortion\n",
    "    \n",
    "    #Re-discretize the data by adding back the mean image, discretizing then subtracting mean\n",
    "    distorted_data += meandisc_image\n",
    "    distorted_data = (distorted_data//(256/pts))*(256/pts) + 0.5*(256/pts)\n",
    "    distorted_data -= meandisc_image\n",
    "    \n",
    "    for j in range(100):\n",
    "        low = 100*j\n",
    "        up = 100*(j+1)\n",
    "        net.blobs['data'].data[...] = distorted_data[low:up]\n",
    "        output = net.forward()\n",
    "        backward = net.backward(**{net.outputs[0]: net.blobs['loss'].data})\n",
    "        distortion[low:up] = np.sign(backward['data'])\n",
    "        y_pred[low:up]=np.argmax(output['loss'],axis =1)\n",
    "    \n",
    "    epsilonvec[i] = epsilon\n",
    "    accvec[i,0] = (y_pred == y_test).mean()\n",
    "\n",
    "    print \"distortion: %f, accuracy: %f\"% (epsilon,(y_pred == y_test).mean())\n",
    "    \n",
    "np.save(\"accuracy_evolution\", accvec)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAegAAAFVCAYAAAAkBHynAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12c2eda50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVMAAADpCAYAAACZbdTTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nXQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12dfb4790>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Plot the results of these tests\n",
    "\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "accvec = np.load(\"numpy_data/accuracy_evolution.npy\")\n",
    "epsilonvec = np.array([256*0.01*i for i in range(40)])\n",
    "width = 2\n",
    "height = 1.25\n",
    "fig, ax1 = sns.plt.subplots()\n",
    "plt.figure(num=None, figsize=(5, 3.2), dpi=80, facecolor='w', edgecolor='k')\n",
    "sns.plt.plot(epsilonvec/256,accvec[:,0],'o-',epsilonvec/256, accvec[:,1],'o-',epsilonvec/256, accvec[:,2],'o-')\n",
    "sns.plt.xlabel(\"Distortion Strength, \"+ r\"$\\epsilon$\")\n",
    "sns.plt.ylabel(\"Test Accuracy\")\n",
    "sns.plt.legend([r\"$2^{2}$\", r\"$2^3$\", r\"$2^4$\"])\n",
    "sns.plt.ylim((0,1))\n",
    "sns.plt.xlim((0,0.25))\n",
    "#plt.legend([\"Raw\", \"GrayScale\", \"Discretized\"])\n",
    "sns.plt.savefig(\"accuracies_evolution1.eps\",figsize=(4,3))\n",
    "sns.plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAegAAAFVCAYAAAAkBHynAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12d57a810>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVMAAADpCAYAAACZbdTTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12c7d3610>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#More results plotting\n",
    "\n",
    "import seaborn as sns\n",
    "#rawdata['X_train'],rawdata['y_train'],rawdata['X_test'],rawdata['y_test'] = load_CIFAR10('cs231n/datasets/cifar-10-batches-py')\n",
    "flatui = [ \"#e74c3c\", \"#34495e\", \"#2ecc71\"]\n",
    "colors = [\"amber\", \"faded green\", \"dusty purple\"]\n",
    "sns.set_palette(sns.xkcd_palette(colors))\n",
    "\n",
    "width = 2\n",
    "height = 1.25\n",
    "\n",
    "fig, ax = sns.plt.subplots()\n",
    "plt.figure(num=None, figsize=(5, 3.2), dpi=80, facecolor='w', edgecolor='k')\n",
    "sns.plt.plot(epsilonvec/256,accvec[:,3],'o-',epsilonvec/256, accvec[:,4],'o-',epsilonvec/256, accvec[:,5],'o-')\n",
    "sns.plt.xlabel(\"Distortion Strength, \"+ r\"$\\epsilon$\")\n",
    "sns.plt.ylabel(\"Test Accuracy\")\n",
    "sns.plt.legend([r\"$2^{5}$\", r\"$2^6$\", \"Original, \" + r\"$2^8$\"])\n",
    "sns.plt.ylim((0,1))\n",
    "sns.plt.xlim((0,0.25))\n",
    "#plt.legend([\"Raw\", \"GrayScale\", \"Discretized\"])\n",
    "sns.plt.savefig(\"accuracies_evolution2.eps\",figsize=(4,3))\n",
    "sns.plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Accuracies of different models tested against eachother"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_test:  (10000, 3, 32, 32)\n",
      "X_train:  (50000, 3, 32, 32)\n",
      "y_train:  (50000,)\n",
      "y_test:  (10000,)\n"
     ]
    }
   ],
   "source": [
    "#Re-load the data\n",
    "rawdata={}\n",
    "\n",
    "rawdata['X_train']=rawdata['X_train'].transpose(0,3,1,2) \n",
    "rawdata['X_test']=rawdata['X_test'].transpose(0,3,1,2) \n",
    "\n",
    "for k, v in rawdata.iteritems():\n",
    "    print '%s: ' % k, v.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy=  0.6011\n"
     ]
    }
   ],
   "source": [
    "#Load the model\n",
    "\n",
    "model_path = '/Users/amaharaj/Documents/Courses/neural_networks/project/assignment2/' # substitute your path here\n",
    "net_fn   = model_path + 'cifar10_deploy.prototxt'\n",
    "param_fn = model_path + 'cifar10_disc128.caffemodel.h5'\n",
    "\n",
    "# Patching model to be able to compute gradients.\n",
    "# Note that you can also manually add \"force_backward: true\" line to \"deploy.prototxt\".\n",
    "\n",
    "net = caffe.Classifier(net_fn, param_fn,\n",
    "                       channel_swap = (2,1,0)) # the reference model has channels in BGR order instead of RGB\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "#the number of discretizations\n",
    "pts = 4\n",
    "tdata = {}\n",
    "tdata['X_train_disc'] = (rawdata['X_train']//(256/pts))*(256/pts) + 0.5*(256/pts)\n",
    "tdata['X_test_disc'] = (rawdata['X_test']//(256/pts))*(256/pts) + 0.5*(256/pts)\n",
    "tdata['X_train_disc'][tdata['X_train_disc'] > 255.0] = 255.0\n",
    "tdata['X_test_disc'][tdata['X_test_disc'] > 255.0] = 255.0\n",
    "\n",
    "# Subtract means from the data\n",
    "meandisc_image = np.mean(tdata['X_train_disc'], axis=0)\n",
    "mean_image = np.mean(rawdata['X_train'], axis=0)\n",
    "tdata['X_train_disc']-= meandisc_image\n",
    "tdata['X_test_disc']-= meandisc_image\n",
    "\n",
    "y_pred = np.zeros(10000)\n",
    "y_test = rawdata['y_test']\n",
    "\n",
    "for j in range(100):\n",
    "    low = 100*j\n",
    "    up = 100*(j+1)\n",
    "    net.blobs['data'].data[...] = tdata['X_test_disc'][low:up]\n",
    "    out = net.forward()\n",
    "    y_pred[low:up]=np.argmax(out['loss'],axis =1)\n",
    "\n",
    "\n",
    "print \"accuracy= \",(y_pred == y_test).mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAcoAAAFgCAYAAAA7GzlUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n+D8HJCcCbmwlaQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x703253690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "vals = [\"4\",\"8\",\"16\",\"32\",\"64\",\"128\",\"256\"]\n",
    "\n",
    "\n",
    "heatmap_arr = np.array((7,7))\n",
    "heatmap_arr= np.array([[0.7517,0.7478,0.7348,0.729,0.7284,0.7277,0.7279]\n",
    ",[0.7079,0.7845,0.7847,0.7777,0.7737,0.7739,0.7734]\n",
    ",[0.6695,0.7703,0.7977,0.8001,0.7982,0.7978,0.7982]\n",
    ",[0.6294,0.7452,0.7949,0.8034,0.8028,0.8037,0.8039]\n",
    ",[0.6320,0.7400,0.7915,0.8076,0.8078,0.8063,0.8064]\n",
    ",[0.6011,0.7120,0.7838,0.8026,0.8040,0.8037,0.8032]\n",
    ",[0.6532,0.7423,0.7942,0.8042,0.8069,0.8077,0.8074]])\n",
    "\n",
    "import seaborn as sns\n",
    "sns.heatmap(heatmap_arr*100,annot=True, fmt=\".2f\", linewidths=.9, xticklabels=vals,yticklabels=vals)\n",
    "sns.plt.xlabel(\"Discretization of test set\")\n",
    "sns.plt.ylabel(\"Discretization of training set\")\n",
    "\n",
    "sns.plt.savefig(\"heatmap_accuracies.eps\")\n",
    "np.save(\"accuracy_heatmap\",heatmap_arr)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}