4 ms·
Do they have a model of your face already trained, or are they just using a generic facial recognition neural network to compute an embedding vector and then fi
by cycrutchfield 9y ago
Do they have a model of your face already trained, or are they just using a generic facial recognition neural network to compute an embedding vector and then finding identities that are close to that vector in L2 distance or whatever?
- bjt2n3904 9y agoHow do they discern the difference between someone without an account (ie: a new face), a poor result from the network, and someone who's opted out? Unless the default for everyone is "opt-out", which I highly doubt.
- cycrutchfield 9y agoMaybe they use some sort of locality-sensitive hash to efficiently gather a set of (approximate) candidates that are close by in that embedding space, and then use some heuristics to further narrow down who the most probable match is.