Image recognition uses deep learning, and in particular CNNs to train on and recognise faces. Usually, this entails training on lots of data. However, recently, we have seen face recognition being deployed everywhere, and being used for passport control, for example, and some airlines have deployed in lieu of boarding pass scanners. How is this accomplished? How can they achieve such accuracy without having hundreds of pictures of everyone? How can, for example, governments, with a quick scan of my face, recognise me and identify my name and identity without having any picture of me besides the ones available from my official ids? There is a link here which, however, does not really explain it technically. Is it done using neural nets? What is their architecture. Are there papers describing it (technically) anywhere?
-
$\begingroup$ good question... I think one shot learning plays an important role $\endgroup$– PeterJan 3, 2020 at 17:31
-
$\begingroup$ I am afraid that not revealing how it's done is a big security feature. Also we don't have any performance measure. I suspect they are not that well performing... -most of the work could even be done by a human remotly without us knowing- They may even just mostly work by prevention, like a fake camera would prevent robberies, even without anyone looking at the pictures. $\endgroup$– Lucas MorinJan 5, 2020 at 11:17
1 Answer
Have a look at a few networks: -Siamese Network -One-Shot Learning Model
Once going through these models you will understand how they can very well with very limited amounts of data.