4 ms·
If some kind of facial recognition algorithm/model can use worse inputs to "do better than humans" in some task (i.e. try to recognize between twins using a dow
by mrbungie 3y ago
If some kind of facial recognition algorithm/model can use worse inputs to "do better than humans" in some task (i.e. try to recognize between twins using a downscaled image from a 720p laptop camera) more often than not focusing in other things different from the face itself or there is some kind of target leakage in the datasets. More often than not those are big reasons why is so difficult to target SOTA paper performance in the real world, be it in face recognition or otherwise.
ML algorithms for FR are not magical and I don't expect them to be always better in all conditions than a human with two eyes (i.e. stereo vision).
- fwlr 3y agoI won’t say it’s impossible, but your suggestions do seem unlikely to me. We’re talking about unlabeled data, cropped to just the face, and the state of the art being found in commercial models in the real world rather than in papers competing to pass standard tests in the literature. (From my direct experience they arguably are pretty close to magical; I think there is just a lot more information in pictures of faces than we thought there was. It’s quite plausible that, since native human machinery was evolved in groups of maximum ~150 distinct faces and was optimized for things like detecting minor emotional cues, our brains are simply throwing a lot of that information out. But that’s not really the point I want to make here; instead, I want to say stuff like “commercial models aren’t trailing research papers; research papers are trailing commercial models”.)