LogInTake a look at this interesting face-aging GAN created by L'Oréal, a French personal care and make-up company.
Last year, at ECCV 2020, L'Oréal presented AgingMapGAN (AMGAN), a high-resolution controllable face aging neural network with spatially-aware conditional GANs. Unlike existing approaches and datasets for face-aging that produce results skewed towards the mean, with individual variations and expression wrinkles often invisible or overlooked in favor of global patterns such as the fattening of the face, this GAN changes the appearance of a high-resolution image using ethnicity-specific aging information and weak spatial supervision to guide the aging process.
"The model takes a patch p from the input image I, a target aging map A, and two orthogonal gradient images X and Y. The image patch Ip is then transformed according to the local aging information contained in the map Ap, while the orthogonal gradients Xp and Yp provide the coordinates of the patch in a fully convolutional manner. The conditions are injected into the generator via the SPADE block to preserve spatial information. Finally, the generator uses an attention mechanism to only change relevant parts of the image, thus preserving the clothes, earrings, and other facial features unrelated to aging," reads the project paper.
Learn more about this GAN here. Also, don't forget to join our new Reddit page, our new Telegram channel, follow us on Instagram and Twitter, where we are sharing breakdowns, the latest news, awesome artworks, and more.