Research map / 3D vision / 3D generation
3D CoMPaT: Composition of Materials on Parts of 3D Things
Yuchen Li, Ujjwal Upadhyay, Habib Slim, Tezuesh Varshney, Ahmed Abdelreheem, Arpit Prajapati, Suhail Pothigara, Peter Wonka, Mohamed Elhoseiny
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 a separate 2D shape classifier to 2D GCR-SEG (BPNet) | 2D GCR-SEG (BPNet) · 3D CoMPaT GCR (Top-1 predicted shape) | Value-all: 3.54 -> 16.57 | Improved |
| add a separate 3D shape classifier (PCT) to 3D GCR-SEG (BPNet) | 3D GCR-SEG (BPNet) · 3D CoMPaT GCR (Top-1 predicted shape) | Value-all: 2.02 -> 8.12 | Improved |
| add a separate 3D shape classifier to 3D GCR-SEG (BPNet) | 3D GCR-SEG (BPNet) · 3D CoMPaT GCR (Top-1 predicted shape) | Shape accuracy: 36.91 -> 67.53 | Improved |
| provide ground truth parts instead of predicted parts (ablation) | 2D GCR-SEG (BPNet) + separate 2D shape classifier · 3D CoMPaT GCR (Top-1 predicted shape) | Value-all: 16.57 -> 46.58 | Improved |
| provide ground truth materials instead of predicted materials (ablation) | 2D GCR-SEG (BPNet) + separate 2D shape classifier · 3D CoMPaT GCR (Top-1 predicted shape) | Value: 40.39 -> 42.98 | Improved |
| provide ground truth materials instead of predicted materials (ablation) | 2D GCR-SEG (BPNet) + separate 2D shape classifier · 3D CoMPaT GCR (Top-1 predicted shape) | Value-all: 16.57 -> 16.57 | No clear effect |
| provide ground truth parts instead of predicted parts (ablation) | 3D GCR-SEG (BPNet) + separate 3D shape classifier · 3D CoMPaT GCR (Top-1 predicted shape) | Value-all: 8.12 -> 42.42 | Improved |
| train PointMLP on 3DCoMPaT (one composition per shape) instead of ModelNet40 | PointMLP · ScanObjectNN hardest variant (9 classes) | accuracy: 24.33 -> 29.21 | Improved |
| replace PointNet++ with PCT for 3D part classification | 3D part classification · 3D CoMPaT 3D part classification | accuracy: 24.18% -> 37.37% | Improved |
| replace PointNet++ with PCT for 3D shape classification | 3D shape classification · 3D CoMPaT 3D shape classification | accuracy: 57.95% -> 69.09% | Improved |
Closest papers by meaning
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.93)
- Common3D: Self-Supervised Learning of 3D Morphable Models for Common Objects in Neural Feature Space (CVPR 2025 · similarity 0.87)
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.87)
- AutoPartGen: Autoregressive 3D Part Generation and Discovery (NeurIPS 2025 · similarity 0.87)
- Structured 3D Latents for Scalable and Versatile 3D Generation (CVPR 2025 · similarity 0.87)
- Large-Vocabulary 3D Diffusion Model with Transformer (ICLR 2024 · similarity 0.86)
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.
