TPG1 – Discussion
March 22nd, 2007 by eleanor - TPG1
Blog Mentions and Comments
“see what you mean” 4.18.07 on ads without products
“Internet art made from facial-recognition app’s Flickr mistakes” 4.4.07 on BoingBoing
“Flickr Bots Eats Facebook” 4.4.07 on New Visuals
“Stories about faces only robots can see” 4.4.07 on Techslut
4.3.07 on networked_performance
Please feel free to use this space to discuss “Anthroptic” and any other related topic.
5 Comments »
Additional comments powered by BackType
I saw this (fairly bad) movie the other night, Deja Vu, with Denzel Washington, where they used facial-recognition software on a backpack. art meets other art!
Check out this new commercial for a sony camera with facial recognition technology that finds the faces in a picture “because,” as a voice-over says, “the face makes the photo.” Its even got little green boxes and everything.
In defense of the algorithm, none of the photos were false-matches to my own face (there has been confusion about that)… simply erroneously detected as human. And actually, for three of the eight photos Ben & I stretched our definition of what counts as a match to a non-face (we included a match to a stencil of a human face, a match to a photo within the photo, and a match to a kitten). Also, keep in mind that there is an issue of scale here–the algorithm has so far evaluated 2,250,000+ photos–so even if an error is very unlikely, it will happen.
Finally, as explained by TPG, I did adjust the settings on the algorithm. When using facial-recognition software you have to balance the tendency false false-positives (i.e., not identifying a face that is there and/or not recognizing a match) vs. false positives (i.e., identifying a non-existent face and/or match). I turned the “knob” to err on the side of false positives.
wow, your face recognition algorithm must suck!
As part of the “Self-Portrait” project, Ethan “loosened” the parameters in order to produce more results.