Variational Autoencoder for Bollywood actors face image generation implemented with PyTorch.
There is no publicly available Bollywood actors' face dataset, so I scrapped them from Google Image using Selenium. Faces from the images were extracted using dlib.
Generated images are far from ideal, mainly because input data was quite dirty as I didn't have time to clean the data.
Training images (original vs. reconstruction)
Testing images (original vs. reconstruction)
New Samples