Files related to Joel Pitt's Kiwicon 2015 talk on facial recognition and detection technology

image-credits.md

Image Credits

• Tensegrity photo by James Myers

• Multi camera mount: https://www.itech-ny.com/blog/bid/283101/Can-I-Fast-Forward-Through-My-Surveillance-Camera-System-Footage

• NZ Herald article on Auckland CCTV network: http://m.nzherald.co.nz/nz/news/article.cfm?c_id=1&objectid=11550646

• Example surveillance camera photo: http://www.silive.com/news/index.ssf/2009/03/staten_island_stores_surveilla.html

• NeuralTalk and Walk https://vimeo.com/146492001

• Yoda riding a skateboard: https://medium.com/@samim/generating-captions-c31f00e8396e

• Lena https://en.wikipedia.org/wiki/Lenna

• Fabio Lanzoni first used by Needell and Ward in http://arxiv.org/abs/1202.6429

• Example images on recognition pages generated from examples in Kaggle's keypoint detection competition: https://www.kaggle.com/c/facial-keypoints-detection

• Ironsword image https://en.wikipedia.org/wiki/Ironsword:_Wizards_%26_Warriors_II

• Labelled Faces in the Wild http://vis-www.cs.umass.edu/lfw/results.html

• Haar-Cascade Classifier by Viola and Jones (2001) https://www.cs.cmu.edu/~efros/courses/LBMV07/Papers/viola-cvpr-01.pdf

• OpenCV docs on Haar-Cascade Classifier http://docs.opencv.org/2.4/modules/objdetect/doc/cascade_classification.html

• Label spam rain forest https://commons.wikimedia.org/wiki/File:Rain_forest_NZ.JPG

• Label spam by deep dream https://github.com/burningion/deepgraffiti

• Prosthetic face mask http://www.urmesurveillance.com/urme-prosthetic/

• Outlines of faces at different angles http://www.newtemplate13.tk/drawing-faces/

• Mask ad: http://www.aliexpress.com/item/Wholesale-China-No-1-Famous-Brand-CM-Gauze-Masks-for-PM2-5-Air-Pollution-Anti-Dust/2047359734.html

• Passengers wearing masks on train https://en.wikipedia.org/wiki/File:Swine_Flu_Masked_Train_Passengers_in_Mexico_City.jpg

• Masks used during clashes with police in Hong Kong http://www.theguardian.com/world/2014/oct/13/hong-kong-barricades-rushed-large-crowd

• Mask examples: http://www.ebay.com/bhp/ghost-mask https://commons.wikimedia.org/wiki/File:GuyFawkesMask.jpg https://en.wikipedia.org/wiki/Mexican_mask-folk_art#/media/File:DevMaskPasMAPDF.JPG

• Religion/Cultural face clothing:

  • Afghanistan Burqa https://commons.wikimedia.org/wiki/File:Burqa_Afghanistan_01.jpg
  • Niqab https://commons.wikimedia.org/wiki/Niqab#/media/File:EFatima_in_UAE_with_niqab.jpg

• Smog masks in China http://gbtimes.com/china/photos-chinese-wear-cute-face-masks-fight-air-pollution

• CV Dazzle http://cvdazzle.com/

• Dazzle Camo https://en.wikipedia.org/wiki/File:USS_West_Mahomet_(ID-3681)_cropped.jpg https://en.wikipedia.org/wiki/Dazzle_camouflage#/media/File:HMS_Argus_(1917)_cropped.jpg

• AVG Invisibility Glasses http://now.avg.com/avg-reveals-invisibility-glasses-at-pepcom-barcelona/

• IR LED array and security camera https://en.wikipedia.org/wiki/Facial_recognition_system#/media/File:Surveillance_equipment_5413.jpg

lfw_performance.py

# Graph LFW performance through time

from pandas.io.parsers import read_csv
from pandas import DataFrame

import matplotlib.pyplot as plt
import numpy as np

df = read_csv('lfw_through_time.csv', skipinitialspace=True)

data = df[['year', 'accuracy', 'externaldata', 'commercial']]
have_years = data.dropna(subset=['year'])

no_external_data = have_years.loc[df['commercial'].isnull()].loc[df['externaldata'] == 0]
w_external_data_not_commercial = have_years.loc[df['commercial'].isnull()].loc[df['externaldata'] == 1]
w_external_data_commercial = have_years.loc[df['commercial'] == ' Commercial'].loc[df['externaldata'] == 1]
w_external_data_commercial = have_years.loc[df['commercial'] == 'Commercial ']


def plot_graph(fn, datasets):
    plt.figure(figsize=(6.5, 6.5))
    plt.style.use('dark_background')
    ax = plt.subplot(111)
    ax.grid(True)
    ax.set_xlim([1990,2016])
    ax.set_ylim([0.0,1.0])
    ax.set_ylabel('Accuracy')
    ax.set_title('LFW performance by publication year')
    gridlines = ax.get_xgridlines() + ax.get_ygridlines()
    for line in gridlines:
        #line.set_color('white')
        line.set_linestyle(':')
    for data, marker, c in datasets:
        ax.scatter(data['year'], data['accuracy'], marker='o', s=100, c=c, edgecolors='face')
    plt.savefig(fn, transparent=True)

plot_graph("1.png", [(no_external_data, 'v', 'green')])
plot_graph("2.png", [(no_external_data, 'v', 'green'), (w_external_data_not_commercial, 's', 'blue')])
plot_graph("3.png", [(no_external_data, 'v', 'green'), (w_external_data_not_commercial, 's', 'blue'), (w_external_data_commercial, 'o', 'red')])

lfw_through_time.csv