[timdumol]

Actually, based on my reading of the paper, it seems that they learn a representation using one data set (the one with 1000 labeled samples per identity), then use that representation to classify on other training sets (like the Labeled Faces in the Wild, which has 13,323 photos of 5,749 celebrities). In fact, from what I can tell from section 5.1, they seemed to use face pairs (and so trained on 1 sample per person, and then tested on the other sample).

tl;dr: They don't need 1,000 labeled samples per identity (once done with the representation phase), and they achieved 97.25% accuracy on ~6,000 distinct identities, with only one training photo per identity.

[thatcherclay]

I think you guys are reading different parts of the article.

They present both results, supervised and unsupervised (where unsupervised uses the SFC dataset to train). They achieved 95.92% accuracy LFW with unsupervised (section 5.3) - so they can train on SFC and then classify a single image in a different domain with 95.92% accuracy.

They achieved the 97.25% accuracy level was achieved as you say, when they let the pairs into the training set. But they overfit with LFW alone, and has to add an additional 100k identities with more samples (30) per identity. A very impressive measure, but not quite as good as being able to generalize with 97% accuracy from a single training photo.

[JoeAltmaier]

I looked hard, and could not prove they can identify a single image with that accuracy. I couldn't tell WHAT the trial protocol was. Where is that in the document?

[apu]

The LFW test is verification: given two faces you've never seen before, tell whether they're the same person or not. That's a much weaker test than saying "who is this?"

The LFW benchmark protocol is described in the original tech report: http://vis-www.cs.umass.edu/lfw/lfw.pdf