Reproduce “PartNet: A Large-scale Benchmark for Fine-grained and Hierarchical Part-level 3D Object Understanding” 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
Kaichun Mo, Shilin Zhu, Angel X. Chang, Li Yi, Subarna Tripathi, Leonidas J. Guibas, Hao Su
The authors introduce PartNet, a dataset of 26,671 3D CAD shapes across 24 categories, with 573,585 fine-grained, hierarchical part-instance labels. They also propose an instance-segmentation method that predicts semantic labels and candidate masks. On the geometry-based benchmark, it achieves 54.4% Part-category mAP versus SGPN’s 46.8%; evaluation excludes rare and non-geometric parts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.