The technology behind image aging became famous thanks to an application called FaceApp, which went viral across the world. Although there have been different attempts at approaching face-aging in the past, they have confronted limitations like needing a lot of data, producing ghosting artifacts (not looking natural), and an inability to do the inverted operation (revert back from old to young). Simply put, the results fell short of our expectations.
With the recent success of GAN-based architectures, we can now generate high-resolution and natural-looking output. In this tutorial we will train CycleGAN, one of today's most interesting architectures, to do forward aging from 20s to 50s and reverse aging from 50s to 20s.
This notebook has a complementary tutorial, Implementing CycleGAN for Age Conversion.
