Anything-of-anything/Anything-3D

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1633 stars · 82 forks observed · 2026-08-28

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

Anything-3D is a Python research project that combines Meta's Segment Anything model with a series of 3D models (3DFuse, Zero 1-to-3, NeRF, HRN) to lift segmented objects from images into 3D. It provides several demos including 3D object reconstruction, novel view synthesis, NeRF scene reconstruction, and face reconstruction from in-the-wild images.

Use cases

• segment an object from a photo and reconstruct it in 3d
• generate novel views of an object from a single image
• improve nerf reconstruction by removing complex backgrounds
• reconstruct a detailed 3d face from an in-the-wild photo
• experiment with segment anything combined with 3d generation models

When to choose

• you want to combine segment anything with 3d reconstruction or novel view synthesis models
• you need a reference implementation for segmentation-then-3D pipelines
• you are researching object lifting, NeRF preprocessing, or single-image 3D face reconstruction

When to avoid

• you need a production-ready or well-maintained tool with ongoing support
• you lack a GPU or cannot set up heavy deep learning dependencies
• you need a polished end-user application rather than research demo code

Facets

library · maturity experimental

computer-vision image-processing machine-learning graphics computer-vision image-processing deep-learning graphics python segment-anything 3d-reconstruction nerf novel-view-synthesis face-reconstruction research-demo linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Anything-of-anything/Anything-3D")