notebooks/20200106 Analyse groundtruth results.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Analyse groundtruth results\n",
    "An example of how to read in results and output the labelled data, some statistics and visuals about categories."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import os\n",
    "import glob\n",
    "import logging\n",
    "from collections import Counter, defaultdict\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.append(os.path.abspath('../'))\n",
    "from gandt.data import analyse_labels_data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "These files should be copied from S3, e.g. using aws CLI:\n",
    "\n",
    "```aws s3 sync s3://your-bucketk/20200103-wedding-user/labelled/20200103-wedding-user/annotations ../data/external/labelled/20200103-wedding-user/annotations/```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "annotation_files = glob.glob('../data/external/labelled/20200103-*/annotations/consolidated-annotation/consolidation-request/iteration-1/*.json')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Read in responses and preprocess"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw_responses = pd.concat([analyse_labels_data.read_responses(x) for x in annotation_files])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sometimes there are duplicates in label files, so get rid of them, only accept the first response per worker per image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw_responses = raw_responses.groupby(['worker_id', 'image_filename', 'image_index'], as_index=False).first()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Filter out responses from unreliable workers. `unfiltered_responses` contains data from all workers (useful for debugging), `filtered_responses` contains only data from workers deemed reliable."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "unfiltered_responses, filtered_responses = analyse_labels_data.filter_out_unreliable_workers(raw_responses, reliability_threshold=0.8, min_ratings_no=2, min_ratings_prop=0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Select responses which have sufficient agreement from labellers, and ones which need to be relabelled"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "certain_responses = filtered_responses[filtered_responses['is_label_certain']]\n",
    "\n",
    "responses_to_relabel = filtered_responses[~filtered_responses['is_label_certain']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Select only colums which will be useful for training a model, e.g. filename, majority label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "certain_responses = certain_responses[['image_filename', 'majority_label']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualise some results\n",
    "NB the colour below are just an example that works well for this case (3 binary labels)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0, 'Proportion of majority response [%]')"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "darkred = (0.8, 0.1, 0.1, 1)\n",
    "lightred = darkred[:3] + (0.6,)\n",
    "darkgrey = (0.7, 0.7, 0.7, 1)\n",
    "lightgrey = darkgrey[:3] + (0.6,)\n",
    "darkgreen = (0.3, 0.6, 0.2, 1)\n",
    "lightgreen = darkgreen[:3] + (0.6,)\n",
    "darkblue = (0.2, 0.4, 0.7, 1)\n",
    "lightblue = darkblue[:3] + (0.6,)\n",
    "\n",
    "(filtered_responses.groupby(['labelling_job_name'])['majority_label_description'].value_counts(normalize=True) * 100).sort_index().unstack(1).plot.barh(stacked=True, color=[\n",
    "    darkred, lightred,\n",
    "    darkgrey, lightgrey,\n",
    "    darkgreen, lightgreen,\n",
    "    darkblue, lightblue\n",
    "])\n",
    "plt.legend(bbox_to_anchor=(1.04,1), loc=\"upper left\")\n",
    "plt.ylabel('Type of labelling job')\n",
    "plt.xlabel('Proportion of majority response [%]')"
   ]
  }
 ],
 "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": "3.7.0"
  },
  "pycharm": {
   "stem_cell": {
    "cell_type": "raw",
    "source": [],
    "metadata": {
     "collapsed": false
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}