Meta SAM 3D
Meta's SAM 3D series — extract object/human 3D models from a single image
Meta's SAM 3D series extracts object and human 3D models from a single image.
Image generation and computer vision — detection, editing, 3D.
Key features
- Promptable model for single-image full-body 3D human mesh recovery (HMR)
- Uses Momentum Human Rig (MHR) parametric mesh decoupling skeletal structure and surface
- Supports auxiliary prompts: 2D keypoints and masks, like the SAM family
- State-of-the-art accuracy in in-the-wild conditions, handling occlusions and viewpoints
- Checkpoints and dataset on Hugging Face; web demo and demo.py included
Good for
- Full-body 3D human mesh recovery from a single image
- Applications needing accurate body, feet, and hand pose estimation
Stack & tags
PythonPyTorch3d-human-meshhmrpose-estimationsamopen-weightsmetacomputer-vision
Quick start
hf download facebook/sam-3d-body-dinov3 --local-dir checkpoints/sam-3d-body-dinov3
python demo.py --image_folder <path_to_images> --output_folder <path_to_output> --checkpoint_path ./checkpoints/sam-3d-body-dinov3/model.ckpt --mhr_path ./checkpoints/sam-3d-body-dinov3/assets/mhr_model.ptFrom the project README on GitHub.
Similar open-source projects
FLUX.2 klein
★ 25,996Black Forest Labs' sub-second image generation — generate and edit in <0.5s
DiffSynth-Studio
★ 13,200The framework behind Qwen-Image-i2L — the first "Image→LoRA" visual generation system