Research map / Point clouds / Point cloud classification
Primary Visual Cortex Inspired Point Cloud Analysis Framework
Jisheng Dang, Delin Deng, Bimei Wang, Jingze Wu, Hui Zhang, Haijiang Li, Jingmei Jiao, Dengyue Pan, Mangang Xie, Jizhao Liu
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 |
|---|---|---|---|
| replace MLP-based Spiking PointNet with DC-CCNN | point cloud classifier · ModelNet40 | Overall accuracy (%): 83.5 -> 93.5 | Improved |
| use DC-CCNN instead of PointMLP on clean data | DC-CCNN · ScanObjectNN | Overall accuracy (%): 85.4 -> 87.3 | Improved |
| use DC-CCNN instead of PointMLP on clean ModelNet40 | DC-CCNN · ModelNet40 | Overall accuracy (%): 93.2 -> 93.5 | No clear effect |
| use DC-CCNN instead of PointMLP under extreme sparsity | DC-CCNN · ModelNet40 (128 points) | Overall accuracy (%): 37.8 -> 60.3 | Improved |
| use DC-CCNN instead of PointMLP under heavy Gaussian noise | DC-CCNN · ModelNet40 (Gaussian noise 0.09) | Overall accuracy (%): 79.7 -> 90.8 | Improved |
| use DC-CCNN instead of PointMLP under 50% salt-and-pepper noise | DC-CCNN · ModelNet40 (salt-and-pepper 0.5) | Overall accuracy (%): 49.2 -> 62.8 | Improved |
| use DC-CCNN instead of PointMLP under extreme sparsity on real scans | DC-CCNN · ScanObjectNN (128 points) | Overall accuracy (%): 39.7 -> 54.8 | Improved |
| use DC-CCNN instead of PointMLP on ShapeNetPart | DC-CCNN · ShapeNetPart | Class mIoU (%): 83.9 -> 84.6 | Improved |
| add spatial coupling and dynamic threshold to signal accumulation (full model vs signal accumulation only) (ablation) | DC-CCNN · ModelNet40 (salt-and-pepper 0.5) | Overall accuracy (%): 53.0 -> 62.8 | Improved |
| add signal accumulation and dynamic threshold to spatial coupling (full vs spatial coupling only) (ablation) | DC-CCNN · ModelNet40 (salt-and-pepper 0.5) | Overall accuracy (%): 66.9 -> 62.8 | Hurt |
| add spatial coupling and dynamic threshold to signal accumulation (full vs signal accumulation only) (ablation) | DC-CCNN · ModelNet40 (128 points) | Overall accuracy (%): 60.9 -> 60.3 | Hurt |
| add the other two modules to dynamic threshold only (full vs dynamic threshold only) (ablation) | DC-CCNN · ModelNet40 (128 points) | Overall accuracy (%): 54.3 -> 60.3 | Improved |
Closest papers by meaning
- Verifiability and Predictability: Interpreting Utilities of Network Architectures for Point Cloud Processing (CVPR 2021 · similarity 0.85)
- Spiking Point Transformer for Point Cloud Classification (AAAI 2025 · similarity 0.84)
- A Network Architecture for Point Cloud Classification via Automatic Depth Images Generation (CVPR 2018 · similarity 0.84)
- Point-to-Spike Residual Learning for Energy-Efficient 3D Point Cloud Classification (AAAI 2024 · similarity 0.84)
- DC-PCN: Point Cloud Completion Network with Dual-Codebook Guided Quantization (AAAI 2025 · similarity 0.84)
- DeepLA-Net: Very Deep Local Aggregation Networks for Point Cloud Analysis (CVPR 2025 · similarity 0.83)
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.
