3 ms·
I was wondering why skype and any other video chat don't do this five years ago, or more. Probably we needed deep learning to do this well.
by scotty79 7y ago
I was wondering why skype and any other video chat don't do this five years ago, or more.
Probably we needed deep learning to do this well.
- ktpsns 7y agoCorrect. Apple certainly "reused" the infrastructure they established with realtime facial recognition and manipulation, all encapsuled in ARKit (https://developer.apple.com/augmented-reality/arkit/ https://developer.apple.com/augmented-reality/arkit/). Furthermore, these compute intensive codes are accelerated with special chips ("AI chip", "Neural engine", etc.). That's actually quite novel stuff which wasn't there five years ago.
- ThePadawan 7y agoSome research achieved basic results in 2012 [0] and spun off into a plugin ("Catch-Eye") [1] which is no longer available (and the website was last updated in 2017). [0] https://cgl.ethz.ch/publications/papers/paperKus12.php https://cgl.ethz.ch/publications/papers/paperKus12.php [1] https://www.catch-eye.com/index.html https://www.catch-eye.com/index.html
- moftz 7y agoYou might also need the FaceID sensor as well. It turns your nose a little bit too apparently so I imagine that it's creating some sort of mask and mapping your facial movements onto it.