README.md

metadata

license: mit

This dataset contains benchmarks for open-world object part segmentation proposed by Find Any Part in 3D (ICCV 2025). This dataset includes two human-annotated benchmarks: Objaverse-General (of 100 object categories) and ShapeNetPart-Objaverse (of the same categories of ShapeNetPart, but with objects source from Objaverse to study distribution shift).

Usage

Inside both objaverse-general and objaverse-shapanetepart directories, the benchmark has the following directory structure: Each object is stored in a folder with the structure below:

{object_name}_{objaverse_uid}/
├── label_map.json # Maps part IDs to natural language part names
├── labels.npy # Per-point label assignment (0 means no label)
└── points5000.pcd # Object point cloud with 5000 points