Research map / Point clouds / Point cloud classification
Spiking PointNet: Spiking Neural Networks for Point Clouds
Dayong Ren, Zhe Ma, Yuanpei Chen, Weihang Peng, Xiaode Liu, Yuhan Zhang, Yufei Guo
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 |
|---|---|---|---|
| train with 1 time step and infer with 4 instead of training with 4 time steps (ablation) | Spiking PointNet · ModelNet40 | accuracy: 86.70% -> 87.13% | Improved |
| train with 1 time step and infer with 4 instead of training with 4 time steps (ablation) | Spiking PointNet · ModelNet10 | accuracy: 91.05% -> 92.32% | Improved |
| infer with multiple time steps instead of 1 (trained 1 T) (ablation) | Spiking PointNet · ModelNet40 | accuracy: 86.98% -> 87.26% | Improved |
| add membrane potential perturbation (ablation) | Spiking PointNet · ModelNet40 | accuracy: 87.13% -> 88.61% | Improved |
| add membrane potential perturbation (ablation) | Spiking PointNet · ModelNet10 | accuracy, test 1 T: 91.99% -> 91.66% | Hurt |
| add membrane potential perturbation (ablation) | Spiking PointNet · ModelNet10 | accuracy, test 4 T: 92.32% -> 93.31% | Improved |
| use surrogate k=5 instead of k=0.5 or k=20 (ablation) | Spiking PointNet · ModelNet40 | accuracy, 1 time step: 80.34% / 83.46% -> 86.98% | Improved |
| increase training time steps from 1 to 4 at k=5 (ablation) | Spiking PointNet · ModelNet40 | accuracy: 86.98% -> 86.70% | No clear effect |
| increase training time steps from 1 to 4 at k=20 (ablation) | Spiking PointNet · ModelNet40 | accuracy: 83.46% -> 75.36% | Hurt |
| average outputs across all time steps instead of using a single step (ablation) | Spiking PointNet (train 4 T) · ModelNet40 | accuracy: 85.70% -> 86.70% | Improved |
| replace PointNet ANN with Spiking PointNet 1 T (MPP) | PointNet · ModelNet10 | energy per forward pass: 1.4 x 10^8 pJ -> 9.2 x 10^6 pJ | Improved |
| replace ANN PointNet with Spiking PointNet 4 T (MPP) | PointNet · ModelNet40 | accuracy: 89.20% -> 88.61% | Hurt |
Closest papers by meaning
- Spiking Point Transformer for Point Cloud Classification (AAAI 2025 · similarity 0.93)
- Point-to-Spike Residual Learning for Energy-Efficient 3D Point Cloud Classification (AAAI 2024 · similarity 0.92)
- Efficient 3D Recognition with Event-driven Spike Sparse Convolution (AAAI 2025 · similarity 0.92)
- Noise-Injected Spiking Graph Convolution for Energy-Efficient 3D Point Cloud Denoising (AAAI 2025 · similarity 0.90)
- Rethinking Spiking Neural Networks from an Ensemble Learning Perspective (ICLR 2025 · similarity 0.89)
- SpikePoint: An Efficient Point-based Spiking Neural Network for Event Cameras Action Recognition (ICLR 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.
