Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

More specifically: Juggalos figured out how to beat facial recognition tech trained on normal faces and trying to recognize normal faces.

While the black and white paint does seem to negate a lot of natural contrast human faces have – and on which the recognition of faces is currently based – there seems a lot of information left to recognize specific faces, especially when considering stereovision or even possible pulsed laser reflection / LIDAR based solutions (think long-range iPhone X sensor).

Even if the recognition of an individual within a database of millions yields too many false positives, the set can be further refined by considering other subject metadata, say geolocation or social connections.

While considering this, it's still important to not succumb to privacy nihilism. Not everything is lost, but it's hard to practice that when 'beating' facial recognition tech means "looking like a fucking juggalo" [1].

[1] https://news.ycombinator.com/item?id=17452411



Speaking of that, I already fed in data from a kinect of the infrared region of interest from a LBP cascade.

It works to identify the perceptual hash of the depth and contour of the face. It beats all makeup, completely.


That wouldn't work against the traditional anti-face detection technology: bandanas.


I'll need some makeup artists from Star Trek TNG to help me get ready in the mornings.


Qa'pla


Fursuit?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: