# wgsxm/PartCrafter

[NeurIPS 2025] PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers

Repository: https://github.com/wgsxm/PartCrafter
Canonical: https://ross.abutalabs.com/products/partcrafter
Homepage: https://wgsxm.github.io/projects/partcrafter/
Language: Python
License: MIT
License Family: permissive
Topics: 3d-generation, 3d-reconstruction, 3d, image-to-3d, 3d-object-generation, 3d-object-reconstruction, 3d-scene-generation, 3d-scene-reconstruction
Last push: 2026-04-16T18:41:43+00:00

## Health v2 (maintenance only)
Score: 53/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 77, release rhythm 35, longevity 32
- inputs: {"age_days": 450, "days_push": 139, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2471, forks 162 (observed 2026-08-28T04:06:54.694709+00:00)

## What it is
PartCrafter is a structured 3D generative model that jointly generates multiple semantically meaningful 3D mesh parts and objects from a single RGB image using compositional latent diffusion transformers. It is the official PyTorch implementation of a NeurIPS 2025 paper, with pretrained checkpoints for both object-level and scene-level generation.

## Use cases
- generate a 3d mesh from a single photo
- image to 3d object reconstruction
- generate 3d models with separate parts from one image
- reconstruct multi-object 3d scenes from an image
- part-aware 3d generation without image segmentation
- create editable 3d assets from pictures

## When to choose
- you need structured, part-level 3D meshes from a single RGB image in one shot
- you want to avoid two-stage segment-then-reconstruct pipelines
- you need both object-level and scene-level 3D generation with pretrained checkpoints
- you want a research-grade, open-source image-to-3D model with training code

## When to avoid
- you need real-time or low-latency 3D generation on CPU-only hardware
- you require production-ready 3D assets with guaranteed watertight, clean topology
- you lack a CUDA GPU or cannot handle large diffusion model checkpoints
- you need text-to-3D generation rather than image-conditioned generation

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, image-processing, graphics, simulation
- domain: artificial-intelligence, computer-vision, graphics, deep-learning, machine-learning
- platform: python, windows
- tags: 3d-generation, image-to-3d, mesh-generation, diffusion-transformer, part-aware-3d, neurips-2025, 3d-reconstruction, gpu, linux, docker

## Member repositories
- wgsxm/PartCrafter (main) score 53

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:54.694709+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:29:09.045630+00:00, confidence not recorded.
  - readme: https://github.com/wgsxm/PartCrafter (fetched 2026-08-28T04:06:54.694709+00:00, sha bd3bae2f9d65)
  - homepage: https://wgsxm.github.io/projects/partcrafter/ (fetched 2026-08-29T10:10:56.137710+00:00, sha 4713468ff10f)
- Data as of 2026-08-30T08:39:29.467469+00:00.