Research map / Point clouds / Point cloud classification
PointCloud Saliency Maps
Tianhang Zheng, Changyou Chen, Junsong Yuan, Bo Li, Kui Ren
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 |
|---|---|---|---|
| drop highest-saliency points instead of random points (200/1024) | PointNet · ModelNet40 | accuracy (%): 87.7 -> 44.3 | Hurt |
| drop highest-saliency points instead of random points (200/1024) | PointNet · 3D-MNIST | accuracy (%): 94.8 -> 49.2 | Hurt |
| use saliency-map drop instead of critical-subset drop | PointNet · ModelNet40 | accuracy (%): 58.1 -> 44.3 | Hurt |
| use saliency-map drop instead of critical-subset drop | PointNet · 3D-MNIST | accuracy (%): 80.0 -> 49.2 | Hurt |
| drop lowest (negative) saliency points | PointNet · ModelNet40 | accuracy (%): 87% ~ 89% -> 91.4% | Improved |
| drop lowest (negative) saliency points | PointNet · 3D-MNIST | accuracy (%): accuracy increases by nearly 1% | Improved |
| drop highest-saliency points | PointNet++ · ModelNet40 | accuracy (%): 88 ~ 90% -> 58.5% | Hurt |
| drop highest-saliency points | DGCNN · ModelNet40 | accuracy (%): 89% ~ 91% -> 64.2% | Hurt |
| drop 600/1024 highest-saliency points rather than random | PointNet · ModelNet40 | accuracy (%): over 80% -> nearly 0 | Hurt |
| add iterative saliency recomputation for high-drop (ablation) | saliency point dropping · ModelNet40 | accuracy degradation | Improved |
| add iterative saliency recomputation for low-drop (ablation) | saliency point dropping · ModelNet40 | accuracy | No clear effect |
| increase scaling factor alpha above 1 (ablation) | saliency point dropping · ModelNet40 | model prediction loss | No clear effect |
Closest papers by meaning
- Salient Object Detection for Point Clouds (ECCV 2022 · similarity 0.89)
- Unsupervised Point Cloud Representation Learning With Deep Neural Networks: A Survey (TPAMI 2023 · similarity 0.89)
- PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation (CVPR 2017 · similarity 0.89)
- Attentional ShapeContextNet for Point Cloud Recognition (CVPR 2018 · similarity 0.89)
- DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis (CVPR 2025 · similarity 0.88)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.87)
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.
