Research map / Point clouds / Point cloud classification
Poly-PC: A Polyhedral Network for Multiple Point Cloud Tasks at Once
Tao Xie, Shiguang Wang, Ke Wang, Linqi Yang, Zhiqiang Jiang, Xingcheng Zhang, Kun Dai, Ruifeng Li, Jian Cheng
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 |
|---|---|---|---|
| Use Poly-PC (base) instead of PointNet++ | Point-cloud shape classification · ModelNet40 | OA: +1.9 | Improved |
| Use Poly-PC (large) instead of PointNet++ | Point-cloud shape classification · ModelNet40 | OA: +3.0 | Improved |
| Use Poly-PC (base) instead of PointNet++ | Point-cloud semantic segmentation · S3DIS area 5 | mIoU: +6.1 | Improved |
| Use Poly-PC (large) instead of PointNet++ | Point-cloud semantic segmentation · S3DIS area 5 | mIoU: +9.1 | Improved |
| Use Poly-PC (base) instead of VoteNet | 3D object detection · SUN RGBD val | mAP@0.25: +3.2 | Improved |
| Add Res-SA to the naive-SA baseline (ablation) | Poly-PC (base) · SUN RGBD val | mAP@0.25: 60.4 -> 60.8 | Improved |
| Add weight-entanglement-based one-shot NAS (ablation) | Poly-PC with Res-SA · SUN RGBD val | mAP@0.25: 60.8 -> 61.6 | Improved |
| Add task-specific attention (ablation) | Poly-PC with Res-SA and NAS · S3DIS area 5 | mIoU: 61.1 -> 61.9 | Improved |
| Add task-prioritization-based gradient balance (ablation) | Poly-PC with Res-SA, NAS, and attention · S3DIS area 5 | mIoU: 61.9 -> 63.0 | Improved |
| Replace naive SA with Res-SA (ablation) | Poly-PC (base) · SUN RGBD val | mAP@0.25: 59.2 -> 62.3 | Improved |
| Replace naive SA with Res-SA (ablation) | Poly-PC (large) · S3DIS area 5 | mIoU: 63.4 -> 66.0 | Improved |
| Use gradient-magnitude homogenization only rather than full TGB (ablation) | Poly-PC (base) · S3DIS area 5 | mIoU: 63.0 -> 61.3 | Hurt |
Closest papers by meaning
- LTA-PCS: Learnable Task-Agnostic Point Cloud Sampling (CVPR 2024 · similarity 0.85)
- MICAS: Multi-grained In-Context Adaptive Sampling for 3D Point Cloud Processing (CVPR 2025 · similarity 0.83)
- DC-PCN: Point Cloud Completion Network with Dual-Codebook Guided Quantization (AAAI 2025 · similarity 0.83)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.83)
- 3D Focusing-and-Matching Network for Multi-Instance Point Cloud Registration (NeurIPS 2024 · similarity 0.82)
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.82)
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.
