Research map / 3D vision / Face and shape
PartCom: Part Composition Learning for 3D Open-Set Recognition
Tingyu Weng, Jun Xiao, Hao Pan, Haiyong Jiang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add part prototypes without constraints (ablation) | GCP baseline · ModelNet 20-20 | OSCR: 85.5 -> 86.6 | Improved |
| add part diversity loss L_pd to part prototypes (ablation) | part prototype model · ModelNet 20-20 | OSCR: 86.6 -> 93.9 | Improved |
| add part balance loss L_pb to part prototypes (ablation) | part prototype model · ModelNet 20-20 | OSCR: 86.6 -> 93.6 | Improved |
| add both L_pd and L_pb (ablation) | part prototype model · ModelNet 20-20 | OSCR: 86.6 -> 95.8 | Improved |
| replace feature concatenation with max pooling (ablation) | PartCom · ModelNet 20-20 | OSCR: 96.7 -> 94.4 | Hurt |
| add PUFS unknown feature synthesis (ablation) | PartCom without PUFS · ModelNet 20-20 | OSCR: 95.8 -> 96.7 | Improved |
| increase number of part prototypes per class from 1 to 3 or more (ablation) | PartCom · ModelNet 20-20 | OSCR | Improved |
| increase cosine similarity threshold delta above 0.1 (ablation) | PartCom · ModelNet 20-20 | OSCR | Hurt |
| replace ARPL with PartCom on confusing mixup tasks | 3D OSR with PointNet++ · ShapeNetMix | OSCR: 58.4 -> 99.1 | Improved |
| use PartCom instead of ARPL | 3D OSR with PointNet++ · ScanObjectNN 5-10 | OSCR: 84.7 -> 97.1 | Improved |
| use class-specific prototype methods (GCP, RPL) instead of softmax | 3D closed-set classifier · ModelNetMix | OSCR: 80.0 -> 63.8 | Hurt |
Closest papers by meaning
- PartField: Learning 3D Feature Fields for Part Segmentation and Beyond (ICCV 2025 · similarity 0.88)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.87)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.86)
- Activating Sparse Part Concepts for 3D Class Incremental Learning (CVPR 2025 · similarity 0.86)
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.86)
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
