3 ms·
After reading the article, I'm still not sure what the purpose of the training is. If they're trying to reconstruct a film from stills, it seems like a failure
by joe_the_user 10y ago
After reading the article, I'm still not sure what the purpose of the training is. If they're trying to reconstruct a film from stills, it seems like a failure since it looks like they wind-up with all sorts of swirly stuff rather than, say, the original film.
If they're trying to create interesting swirly stuff, where do they intend to go after that?
I mean, sure it's aesthetic though not on the level of weirdness of deep dreams modification.
- visarga 10y agoI am not an expert, but: This approach creates powerful embeddings for images. It can convert an image into embedding space and vice-versa, generate images back from embeddings. It is built to function like a perception and imagination module. The embeddings are much lower dimensional and the latent variables are disentangled. There is a component for "has glasses" which you can flip and get the same image + glasses, for example. It is obvious this would be very useful in building all sorts of classifiers, image generators and agents (because agents need to compute reward over state and action space and disentangled representations of the state space are good for this task).
- argonaut 10y agoThe article does not evaluate the quality of the embeddings, so there's not much to say about all of that.