Apple FaceLit creates 3D models from photos. Soon Harry Potter's "live" photos to come true!
Researchers from Apple have developed a generative framework called FaceLit, which can create photorealistic 3D images of faces directly from 2D images without any manual annotation!
To do this, the model uses position and lighting estimators and integrates the Phong model into neural area rendering. The model is trained to generate the shape and properties of the face material to create photorealistic images with multi-projection 3D and lighting matching. This method allows you to create realistic faces with explicit lighting and view control on multiple datasets.
source:https://t.me/hitechdailynews
