ActivityNetData.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/ineschami/emoji/lib/python2.7/site-packages/matplotlib/font_manager.py:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.\n",
      "  warnings.warn('Matplotlib is building the font cache using fc-list. This may take a moment.')\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import numpy as np\n",
    "import seaborn\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data_path = 'data/ActivityNet/activity_net.v1-3.min.json'\n",
    "vid = json.load(open(data_path))['database']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def print_stats(splits=[], return_args=False):\n",
    "    durations = []\n",
    "    nb_prop = []\n",
    "    prop_durations = []\n",
    "    for i, vid_id in enumerate(vid.keys()):\n",
    "        if len(splits)==0: \n",
    "            pass\n",
    "        else:\n",
    "            if vid[vid_id]['subset'] not in splits:\n",
    "                continue\n",
    "        durations.append(vid[vid_id]['duration'])\n",
    "        nb_prop.append(len(vid[vid_id]['annotations']))\n",
    "        for p in vid[vid_id]['annotations']:\n",
    "            prop_durations.append(p['segment'][1]-p['segment'][0])\n",
    "    nb_videos = len(durations)\n",
    "    nb_prop = np.array(nb_prop)\n",
    "    prop_durations = np.array(prop_durations)\n",
    "    if len(splits)==0:\n",
    "        print \"Total number of videos: {}\".format(nb_videos)\n",
    "    else: \n",
    "        print \"Total number of videos in split {}: {}\".format(split, nb_videos)\n",
    "    print \"{}% of videos have no proposals\".format(100.*np.sum(nb_prop==0)/nb_videos)\n",
    "    print \"Average number of proposals per video (excluding videos with no proposals: {}\".format(nb_prop[nb_prop!=0].mean())\n",
    "    print \"Average video duration: {}\".format(np.mean(durations))\n",
    "    print \"Average proposal duration: {}\".format(np.mean(prop_durations))\n",
    "    if return_args: \n",
    "        return durations, nb_prop, prop_durations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of videos: 19994\n",
      "25.2275682705% of videos have no proposals\n",
      "Average number of proposals per video (excluding videos with no proposals: 1.54274247492\n",
      "Average video duration: 116.704419376\n",
      "Average proposal duration: 49.2139085214\n"
     ]
    }
   ],
   "source": [
    "# all splits \n",
    "print_stats()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of videos in split training: 14950\n",
      "0.0% of videos have no proposals\n",
      "Average number of proposals per video (excluding videos with no proposals: 1.54274247492\n",
      "Average video duration: 117.651892308\n",
      "Average proposal duration: 49.2139085214\n"
     ]
    }
   ],
   "source": [
    "durations, nb_prop, prop_durations = print_stats(['training', 'validation'], True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of videos in split training: 5044\n",
      "100.0% of videos have no proposals\n",
      "Average number of proposals per video (excluding videos with no proposals: nan\n",
      "Average video duration: 113.896187748\n",
      "Average proposal duration: nan\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/ineschami/emoji/lib/python2.7/site-packages/ipykernel/__main__.py:23: RuntimeWarning: Mean of empty slice.\n"
     ]
    }
   ],
   "source": [
    "print_stats(['testing'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEZCAYAAAC99aPhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEjMzK8SJxMzMCvl/vLdVEw6YF6AAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10ee2dfd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(durations, bins=100)\n",
    "plt.xlabel('Durations')\n",
    "plt.ylabel('Nb videos')\n",
    "plt.title('Distribution of video duration for train+val splits(in seconds)')\n",
    "plt.axis([0, 400, 0, 2000])\n",
    "plt.grid(True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEZCAYAAAC99aPhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10fc00cd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(prop_durations, bins=100)\n",
    "plt.xlabel('proposal duration')\n",
    "plt.ylabel('Nb videos')\n",
    "plt.title('Distribution of propsals duration for train+val splits (in seconds)')\n",
    "plt.axis([0, 250, 0, 3500])\n",
    "plt.grid(True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def proportion_of_prop_captured(K, prop_durations, c3d_feats_length, iou_thresh=0.5):\n",
    "    # approximation to see if proposal would be captured \n",
    "    max_length = K*c3d_feats_length\n",
    "    return np.mean(iou_thresh*prop_durations<=max_length)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Computing proportion of proposals captured for different values of K\n",
      "K = 16 : 0.296869580298\n",
      "K = 64 : 0.611949358307\n",
      "K = 128 : 0.786203607353\n",
      "K = 256 : 0.951135969476\n",
      "K = 512 : 0.999913284773\n"
     ]
    }
   ],
   "source": [
    "print \"Computing proportion of proposals captured for different values of K\"\n",
    "for K in [16, 64, 128, 256, 512]:\n",
    "    print \"K = {} : {}\".format(K, proportion_of_prop_captured(K, prop_durations, 0.33))"
   ]
  }
 ],
 "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": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}