Show HN: Cafca – a synthetic face dataset with 1.75M images

Category: ai-ml

Tags: synthetic-data, face-dataset, computer-vision, avatars, 3d-reconstruction

Score: 4.5/10 (Innovation: 6, Technical: 4, Documentation: 3, Utility: 5)

Cafca is a synthetic face dataset containing 1.75 million images with multiview diversity, intended for avatar reconstruction, animation, face parsing, and synthetic-to-real transfer learning. The repository itself only hosts the dataset's website and download instructions, not the generation pipeline or tools. It fills a niche for large-scale synthetic face data but lacks technical depth in its codebase.

Target audience: computer vision researchers, ML engineers, graphics developers

Repository: https://syntec-research.github.io/Cafca/cafca_dataset.html · JavaScript · NOASSERTION · 25 stars

View on Hacker News