geopavlakos/hamer

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1124 stars · 165 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

HaMeR (Hand Mesh Recovery) is a transformer-based model that reconstructs 3D hand meshes from single monocular images using the MANO parametric hand model. It is the official research code release for the CVPR 2024 paper 'Reconstructing Hands in 3D with Transformers', including trained models, demo scripts, and an evaluation pipeline.

Use cases

• reconstruct 3d hand meshes from images
• estimate 3d hand pose from a single photo
• run hand pose estimation on video frames
• evaluate hand reconstruction models on benchmarks
• demo 3d hand reconstruction in a browser or colab
• research hand-object interaction reconstruction

When to choose

• you need state-of-the-art monocular 3D hand mesh reconstruction
• your images contain occlusions, truncations, or hands interacting with objects
• you want MANO model parameters as output for downstream 3D work
• you are doing computer vision research on hand pose

When to avoid

• you need real-time full-body pose estimation rather than hands only
• you cannot obtain the MANO model, which requires registration on the MANO website
• you need a production-ready API or service rather than research code
• you lack a GPU, since the ViT-based model is compute-heavy

Facets

library · maturity active

machine-learning deep-learning computer-vision image-processing computer-vision machine-learning deep-learning python hand-mesh-recovery 3d-reconstruction transformer mano pose-estimation research-code cvpr-2024 gpu docker linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="geopavlakos/hamer")