Spiking Point Transformer for Point Cloud Classification

2025-02-19

AI Summary

This paper addresses the longstanding challenge in 3D point cloud classification of simultaneously achieving high computational efficiency and strong performance with Spiking Neural Networks (SNNs). To this end, it proposes the first deep integration of SNNs with the Transformer architecture, yielding a spatiotemporally sparse, energy-efficient model. Key innovations include: (1) queue-driven sampling with direct encoding for adaptive point selection, and (2) a hybrid-dynamics integrate-and-fire (HD-IF) neuron enabling selective neural activation modeling. The method achieves state-of-the-art accuracy on three major SNN-oriented point cloud benchmarks—ModelNet40, ScanObjectNN, and ShapeNetPart. Theoretically, its energy consumption is reduced by at least 6.4× compared to equivalent Artificial Neural Networks (ANNs), demonstrating unprecedented synergy between energy efficiency and classification accuracy.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: 3D Computer Vision
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Data transparency and provenance

Abstract

Spiking Neural Networks (SNNs) offer an attractive and energy-efficient alternative to conventional Artificial Neural Networks (ANNs) due to their sparse binary activation. When SNN meets Transformer, it shows great potential in 2D image processing. However, their application for 3D point cloud remains underexplored. To this end, we present Spiking Point Transformer (SPT), the first transformer-based SNN framework for point cloud classification. Specifically, we first design Queue-Driven Sampling Direct Encoding for point cloud to reduce computational costs while retaining the most effective support points at each time step. We introduce the Hybrid Dynamics Integrate-and-Fire Neuron (HD-IF), designed to simulate selective neuron activation and reduce over-reliance on specific artificial neurons. SPT attains state-of-the-art results on three benchmark datasets that span both real-world and synthetic datasets in the SNN domain. Meanwhile, the theoretical energy consumption of SPT is at least 6.4$ imes$ less than its ANN counterpart.

Problem

• Develops SNN for 3D point cloud classification
• Reduces computational cost in point cloud processing
• Decreases energy use compared to traditional ANNs

Innovation

• Spiking Point Transformer framework
• Queue-Driven Sampling Direct Encoding
• Hybrid Dynamics Integrate-and-Fire Neuron