Check out DiffLocks, appearing at #CVPR2025. From a single image, we estimate about 100K hair strands that you can then physically simulate. We use a dataset of 40K synthetic hair images with ground truth strands. It's all available for research purposes.
Michael J. Black
Meet DiffLocks, one of our CVPR papers: 3D hair from one image or video, instant mesh, AND public 40,000 3D hairstyle dataset!
Learn more about it at our booth 1333 @ #CVPR2025!
Paper link + author names in thread.
#3DBody #SMPL #MachineLearning #HairTech #GenerativeAI @cvprconference.bsky.social
