akanazawa/hmr

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1666 stars · 397 forks observed · 2026-08-28

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

HMR (Human Mesh Recovery) is a TensorFlow implementation of the CVPR 2018 paper 'End-to-end Recovery of Human Shape and Pose', which regresses 3D human body shape and pose (SMPL parameters) directly from a single RGB image. It includes a demo, webcam demo, and training code with pre-trained models.

Use cases

• estimate 3d human pose from a single image
• recover human body shape from photos
• run 3d human mesh reconstruction on webcam feed
• convert images to smpl body model parameters
• train a human mesh recovery model
• use openpose keypoints to crop people for pose estimation

When to choose

• you need 3D human shape and pose from a single RGB image
• you want the original reference implementation of the HMR paper for research
• you need SMPL parameter regression with pre-trained TensorFlow models

When to avoid

• you need a maintained production-ready library (Python 2.7 / TF 1.3 era code)
• you want PyTorch or TF2 support (use community forks instead)
• you need multi-person 3D pose estimation out of the box

Facets

library · maturity maintenance

machine-learning computer-vision image-processing computer-vision machine-learning deep-learning python windows human-pose-estimation human-shape-recovery smpl 3d-human-mesh tensorflow cvpr-2018 research-code linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="akanazawa/hmr")