Hacker News Reader

Comment by stephengillie | Hacker News Reader

> But the most stunning—and creepiest—software developed in Russia is something called FindFace. It’s an app that lets you take a picture of a stranger and then almost instantly, using a facial-recognition algorithm, find the person on a social network. If you’re hoping the software doesn’t work that well, you’ll be disappointed: When I tested the app, it found the right faces all the freaking time. Privacy is so 2015.

Still waiting for social networks to search for you on other social networks, to verify your identify.

We're sorry, this photo doesn't match your face in photos on other social media, so we can't add it to your profile.

View on HN

Important to note: main usage of this app is cyberbullying and chantage.

The English word for the Russian "шантаж" (shantazh; loan-word from the French chantage) is "blackmail".

thank you :)

Any idea how these applications are built and trained?

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?

They could just be using a feature space for face representations that a model is trained to generate given images, which would reduce the identification problem to a n-dimensional clustering problem.

I'm not as familiar with the models in face recognition but something analogous to word2vec may produce reasonable results.

My guess is that they use siamese cnns or autoencoders and look at the latent space. Also, just straight up training a deep neural network on 20 million classes isn't impossible. We've developed an in house technique that greatly accelerates this. Keep your eyes peeled on arxiv for a publication soon ;)

What authors should we be looking for?

Deep, Derp, et al.

Tapabrata Ghosh

Also, the essence of it is sampling the "labels" (the people's face ID) similar to Word2Vec through similarity. When applied to people's face ID, the key phrase is "In feature space, faces of a feather flock together".

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.

Spent a couple of hours of interesting reading. It sounds like faces cluster in high dimensional space. From the perspective of a curious spectator I would love to see a blog post or article that answered:

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?

They are using openface: https://cmusatyalab.github.io/openface/#openface

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.

This reminded me of a news article I read about "eigenfaces" as a facial recognition method way back; here's its corresponding Wikipedia page: https://en.wikipedia.org/wiki/Eigenface

Really, there would be "eigenshapes" for any arbitrary configuration of 3D space. That we can smoothly transition from one face to the next in a higher dimensional matrix is no surprise.

Not much info in the video but they did train for 6 month to get their first good results. There also seem to be neuron combinations for race and eyes...that's pretty much it.

I'd be curious to see whether the network used has similar issues to the ones explored here:

http://www.evolvingai.org/fooling

for recognizing ImageNet images.

I believe this problem is common to many types of machine learning classifiers (not even just neural nets). See the discussion here https://news.ycombinator.com/item?id=8721650

ВКонтакте already has this internal tool

They won't let you update your photo unless it's similar to your photo on other social networks?

Reply on news.ycombinator.com