YuliangXiu/ICON

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1675 stars · 221 forks observed · 2026-08-28

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

ICON is a PyTorch research implementation of a CVPR 2022 method that reconstructs detailed, animatable 3D clothed human avatars from 2D images using implicit functions and SMPL(-X) body priors. It infers front/back clothed-human normals and combines them with local features to produce robust 3D surfaces even for unconstrained poses.

Use cases

• reconstruct a 3D clothed human avatar from a single photo
• generate animatable SMPL-based body meshes from images
• estimate clothed human normals for 3D surface detail
• create virtual avatars for the metaverse or games from photos
• research on implicit-function-based 3D human reconstruction
• extract garment geometry from images for cloth simulation

When to choose

• you need photo-to-3D clothed human reconstruction with SMPL body priors
• you want a research baseline or successor pipeline (ECON) for avatar creation
• you need animatable avatars from unconstrained 2D images

When to avoid

• you need actively developed software - the authors recommend ECON instead
• you lack a GPU or the SMPL model dependencies
• you need production-grade, licensed commercial avatar tooling (license is non-standard)

Facets

library · maturity maintenance

machine-learning computer-vision image-processing deep-learning computer-vision graphics machine-learning artificial-intelligence python cross-platform 3d-reconstruction implicit-functions smpl human-avatar normal-maps pytorch cvpr-2022 pose-estimation mesh-generation virtual-humans gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="YuliangXiu/ICON")