Description
Getting a 3D body pose from one photo fails with occlusions and odd poses using classic methods. SAM 3D Body is Meta's model for recovering a full-body 3D human mesh from a single image.
It accepts point and mask prompts, stays robust to hard poses and occlusion, with checkpoints and notebooks.
Single image:Full mesh.
Prompts:Points and masks.
Robust:Occlusion and poses.
Examples:Notebooks.
It accepts point and mask prompts, stays robust to hard poses and occlusion, with checkpoints and notebooks.
Features
Single image:Full mesh.
Prompts:Points and masks.
Robust:Occlusion and poses.
Examples:Notebooks.
