Reproduce “A Unified Framework for Image-to-3D Part Generation via Variable Granularity” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Qiwen Gu, Xin Zhou, Xiwu Chen, Zhongyang Zhu, Zhuo Zhang, Junqiao Zhao

FlexPart unifies point, box, and mask prompts in one image-to-3D generator for editable 3D parts. It encodes spatial prompts as part conditions and uses boxes and masks as training guidance for sparse points. On PartVerseXL, mask-conditioned Part CD is 0.2938 versus OmniPart’s 0.3283, while joint prompt training improves point-only inference; single-view depth limits leave occluded-part placement difficult.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.