My assumptions are that the process is scraping social networks for tagged pictures; identifying and standardizing faces and then training a CNN network to identify individuals.
The trouble is wrapping my head around how to do that at the scale needed to uniquely identify 20M+ individuals. Do they train the CNNs in smaller batches and run the models in parallel? How do they normalize the results? How do the add incremental data?
The library gets the picture as input and spits out a 128-element vector. The more similar the vectors are for two pictures the more likely they are of the same person.
You don't need to train it, there is a pretrained version available.
Then, use a recurrent ConvNet at the end to pump out probabilities for each class (in this case face) with a low/tractable number of parameters.
What would pairs of maximally distant faces look like?
What would the mediods of a large dataset that was clustered into two clusters look like? Would it be a high level feature that we'd recognize like male or female?
Would faces that were mediods of the cluster be considered more attractive?
Would it be possible to build a similar app that allowed you to identify an individual from memory? Cluster the dataset and present the user with images representative of the mediods of the clusters. Discard images from clusters that weren't selected and re-cluster. Repeat until a handful of images remain.
Could that technique work as a virtual sketch artist and identify a person's doppelganger?
I'm not as familiar with the models in face recognition but something analogous to word2vec may produce reasonable results.