The authors build on recent holistic 3D generation models that use a structured representation anchoring tokens to explicit spatial positions via a coarse mesh (voxel-anchored positional embeddings).
Dual-frame interleaving lets holistic 3D priors produce part-decomposed assets
StructParts uses a segmented mesh scaffold and interleaved attention to inherit geometric detail and structural coherence from pretrained holistic models with lightweight finetuning.
Chinese Tech
Yiqun Zhao · Binbin Huang · Haobin Duan · Zibo Zhao · Shenghua Gao
The University of Hong Kong · Transengram · Tencent · Shenzhen Loop Area Institute
Research Digest··2 min read
The authors introduce StructParts, a framework that reuses holistic 3D generation priors for part-decomposed asset generation.
Why this paper
From Tencent and 3 others
In one line
StructParts reuses a holistic 3D prior via dual-frame coordinate interleaving to generate part-decomposed assets with minimal finetuning.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§