How many faces do you count?

How many faces do you count in this image?

image-ccv-face-mass.png

Perl counts 44:

(3068,1983): 25x25 @ -2.30285501480103
(4442,2028): 25x25 @ -3.287189245224
(2015,2071): 26x26 @ -3.70826029777527
(3458,2041): 31x31 @ 0.825069129467011
(3849,2063): 28x28 @ -2.27478647232056
(3328,2021): 33x33 @ -0.250598877668381
(3413,2024): 34x34 @ 0.513785362243652
(4188,2039): 31x31 @ -0.453457891941071
(2256,2058): 30x30 @ 0.173833101987839
(2809,2110): 31x31 @ -0.524535775184631
(1834,2168): 32x32 @ -1.64895510673523
(2180,2128): 34x34 @ 3.05323696136475
(3222,2446): 33x33 @ -3.49713373184204
(4663,2134): 37x37 @ 3.50697612762451
(3082,2155): 37x37 @ 2.55681991577148
(1651,2125): 36x36 @ 0.686093211174011
(3596,2133): 32x32 @ -2.47773456573486
(1887,2204): 38x38 @ 2.26138424873352
(3733,1876): 34x34 @ -2.05330586433411
(4907,2190): 42x42 @ 6.07879543304443
(4296,2092): 39x39 @ -3.16300702095032
(3188,2126): 39x39 @ -0.357136130332947
(3859,2140): 42x42 @ -0.0315352529287338
(4477,2149): 39x39 @ 3.46311140060425
(3946,2223): 40x40 @ -0.409443378448486
(906,2228): 40x40 @ -0.587313830852509
(3761,2258): 49x49 @ 5.01696395874023
(4978,2280): 46x46 @ -0.811646401882172
(1819,2335): 51x51 @ 6.60061264038086
(4071,2331): 49x49 @ 1.23171555995941
(3430,2230): 46x46 @ -0.422206252813339
(4569,2542): 52x52 @ -3.66873860359192
(5390,2236): 57x57 @ 2.28667879104614
(3980,2381): 54x54 @ -2.23162841796875
(4834,2254): 60x60 @ 0.909286737442017
(2645,2460): 59x59 @ -1.53757762908936
(3394,2470): 65x65 @ 1.81137716770172
(275,2303): 61x61 @ -3.08972954750061
(3159,2462): 80x80 @ -0.73163914680481
(4238,2356): 64x64 @ -1.11184680461884
(4067,2493): 73x73 @ 0.357052862644196
(5187,2235): 91x91 @ -3.84829616546631
(3394,1436): 243x243 @ -0.155338764190674
(189,1099): 915x915 @ -0.393947660923004

In theory, the last number should be some confidence value, but so far, I haven't made much sense of its range.

I hope to get the code packaged and released as Image::CCV before or on the German Perl Workshop 2012 in Erlangen.

6 Comments

Very nice and it detected me too :) But it also matched a foot as a face (right side -- near Liz)

I like how Jesse is mysteriously invisible to the face-recognition algorithm.

Now to use eigenfaces to perform facial recognition between a photo and CPAN avatars! :)

It looks like a third of cpan avatars wouldn't accurately match: http://hexten.net/cpan-faces/

Cool! Me... want... play...

About 50%? But with this image quality it is a pretty good rate.

Leave a comment

About Max Maischein

user-pic I'm the Treasurer for the Frankfurt Perlmongers e.V. . I have organized Perl events including 9 German Perl Workshops and one YAPC::Europe.