MicrostockGroup
Microstock Photography Forum - General => General Stock Discussion => Topic started by: Titus Livius on July 09, 2015, 14:20
-
FAIL !
Google Photos labels black people as 'gorillas'
Google has removed the 'gorilla' tag from its new Photos app, after it was found to be misidentifying images of black people
http://www.telegraph.co.uk/technology/google/11710136/Google-Photos-assigns-gorilla-tag-to-photos-of-black-people.html (http://www.telegraph.co.uk/technology/google/11710136/Google-Photos-assigns-gorilla-tag-to-photos-of-black-people.html)
as expected, their new auto-keyworder algorithm isnt accurate enough and ends up spitting out wrong keywords .. they'll learn a thing or two about stock keywording and why it's still done manually ... from the demos i've seen so far it can't even work with basic subjects let alone complex images.
-
FAIL !
Google Photos labels black people as 'gorillas'
Google has removed the 'gorilla' tag from its new Photos app, after it was found to be misidentifying images of black people
[url]http://www.telegraph.co.uk/technology/google/11710136/Google-Photos-assigns-gorilla-tag-to-photos-of-black-people.html[/url] ([url]http://www.telegraph.co.uk/technology/google/11710136/Google-Photos-assigns-gorilla-tag-to-photos-of-black-people.html[/url])
as expected, their new auto-keyworder algorithm isnt accurate enough and ends up spitting out wrong keywords .. they'll learn a thing or two about stock keywording and why it's still done manually ... from the demos i've seen so far it can't even work with basic subjects let alone complex images.
Brilluant! They have managed to build an algorithm that is both inaccurate and racist at the same time. You could not make this stuff up!
-
http://www.microstockgroup.com/general-photography-discussion/google-apologises-for-photos-app%27s-racist-blunder/msg424638/#msg424638 (http://www.microstockgroup.com/general-photography-discussion/google-apologises-for-photos-app%27s-racist-blunder/msg424638/#msg424638)
-
Brilluant! They have managed to build an algorithm that is both inaccurate and racist at the same time. You could not make this stuff up!
but also with a DSLR it's harder to calibrate the correct exposure with dark skinned subjects, i don't think these face-recognition algos will become perfect, if the face is dark and they dress dark but the rest of the photo is bright there's just too much contrast especially if their face is in shadow.