Hi all, as part of my PhD, I released the world’s largest public dataset of synthetic faces "Cafca".

I used it to train a foundation model for 3D face reconstruction and novel view synthesis. It is also meant to serve as a prior for animatable avatars, semantic segmentation, and relighting. Here is what it contains:

# Images: 1.755 M (512x512) # Identities: 1,500 # Expressions: 13 # Environments: 3 # Views: 30

The dataset comes with GT cameras and pseudo-GT keypoints and 3D meshes (FLAME, via VHAP) and is released under a CC BY-NC-SA 4.0 license.

The easiest way to get started is Hugging Face (see link on dataset page).

Hope you find it useful!