Cafca Synthetic Face Dataset

Cafca is a high-quality synthetic face dataset consisting of 1.75 million images with 1,500 diverse identities, 13 expressions, 3 environments, and 30 views each. It is designed to support research and development in avatar reconstruction, face parsing, synthetic-to-real transfer learning, animation, relighting, camera estimation, and related areas. The dataset includes annotations for semantic regions, camera parameters, and 3D meshes fitted with VHAP.