[P] Fast Face Aging GAN
This project is basically me trying to replicate the faceapp app. Where it can make you look older. This project supports aging a human to different age ranges (can make you younger or older). It also uses identity preserving techniques from the paper “Identity preserved face aging with CGANs” so that the aged face is not too dissimilar from the non-aged face. It’s fast because I benchmarked the model on the fritz benchmarking utility, and the iPhone X runtime comes out to about 30fps on a 512×512 sized image.
You can also train the network with a higher weight on the age loss, to make the effect more drastic. I kept it small to make the effect subtle. From what I have observed, it can be used on non-cropped, raw images of humans in the wild, and it will still do a pretty decent job of aging the face and leaving all other things unchanged (as you can see on the samples on the github page).
For those interested, here’s a link: Fast Face Aging GAN
Suggestions and improvements are more than welcome. There’s also a demo script to try out the pretrained model on your images.