I have used images of fashion models, preprocessed facial part locations, then worked to morph two faces to create a facial template for potential virtual/simulated characters.
Took zaidbhat1234/Image2StyleGAN implementation of the paper "Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space?"
- Denoise the background.
- Facial distortions when facial parts are covered.
- Utilize the segmentation model to filter out the irrelevant background noises.
- Educate the model to reconstruct facial parts in the style of the target image based on their positional and conceptual information.
- Incoporate the concept of "beauty standard" into the encoder. Aim to morph two images in a manner that complements one another.