Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Noise-Injected Spiking Graph Convolution for Energy-Efficient 3D Point Cloud Denoising

Zikuan Li, Qiaoyun Wu, Jialin Zhang, Kaijun Zhang, Jun Wang

TL;DR

This work tackles the energy-intensive regression problem of 3D point cloud denoising by introducing noise-injected spiking graph convolution networks (NI-SGCN). It develops two architectures, NI-PSGCN and NI-HSGCN, that integrate noise-injected spiking neurons with graph convolution to learn robust, disturbance-aware representations while reducing energy consumption. The approach achieves competitive denoising accuracy on PU-Net and PC-Net with substantial energy savings, including a theoretical 12.75x reduction for NI-PSGCN and about 3.9x for NI-HSGCN over a traditional ANN baseline, highlighting the practical viability of neuromorphic computing for 3D data processing. The paper demonstrates the potential of SNNs for energy-efficient 3D sensing pipelines and outlines a path toward deployment on neuromorphic hardware for real-world, low-power 3D data acquisition devices.

Abstract

Spiking neural networks (SNNs), inspired by the spiking computation paradigm of the biological neural systems, have exhibited superior energy efficiency in 2D classification tasks over traditional artificial neural networks (ANNs). However, the regression potential of SNNs has not been well explored, especially in 3D point cloud processing. In this paper, we propose noise-injected spiking graph convolutional networks to leverage the full regression potential of SNNs in 3D point cloud denoising. Specifically, we first emulate the noise-injected neuronal dynamics to build noise-injected spiking neurons. On this basis, we design noise-injected spiking graph convolution for promoting disturbance-aware spiking representation learning on 3D points. Starting from the spiking graph convolution, we build two SNN-based denoising networks. One is a purely spiking graph convolutional network, which achieves low accuracy loss compared with some ANN-based alternatives, while resulting in significantly reduced energy consumption on two benchmark datasets, PU-Net and PC-Net. The other is a hybrid architecture that combines ANN-based learning with a high performance-efficiency trade-off in just a few time steps. Our work lights up SNN's potential for 3D point cloud denoising, injecting new perspectives of exploring the deployment on neuromorphic chips while paving the way for developing energy-efficient 3D data acquisition devices.

Noise-Injected Spiking Graph Convolution for Energy-Efficient 3D Point Cloud Denoising

TL;DR

This work tackles the energy-intensive regression problem of 3D point cloud denoising by introducing noise-injected spiking graph convolution networks (NI-SGCN). It develops two architectures, NI-PSGCN and NI-HSGCN, that integrate noise-injected spiking neurons with graph convolution to learn robust, disturbance-aware representations while reducing energy consumption. The approach achieves competitive denoising accuracy on PU-Net and PC-Net with substantial energy savings, including a theoretical 12.75x reduction for NI-PSGCN and about 3.9x for NI-HSGCN over a traditional ANN baseline, highlighting the practical viability of neuromorphic computing for 3D data processing. The paper demonstrates the potential of SNNs for energy-efficient 3D sensing pipelines and outlines a path toward deployment on neuromorphic hardware for real-world, low-power 3D data acquisition devices.

Abstract

Spiking neural networks (SNNs), inspired by the spiking computation paradigm of the biological neural systems, have exhibited superior energy efficiency in 2D classification tasks over traditional artificial neural networks (ANNs). However, the regression potential of SNNs has not been well explored, especially in 3D point cloud processing. In this paper, we propose noise-injected spiking graph convolutional networks to leverage the full regression potential of SNNs in 3D point cloud denoising. Specifically, we first emulate the noise-injected neuronal dynamics to build noise-injected spiking neurons. On this basis, we design noise-injected spiking graph convolution for promoting disturbance-aware spiking representation learning on 3D points. Starting from the spiking graph convolution, we build two SNN-based denoising networks. One is a purely spiking graph convolutional network, which achieves low accuracy loss compared with some ANN-based alternatives, while resulting in significantly reduced energy consumption on two benchmark datasets, PU-Net and PC-Net. The other is a hybrid architecture that combines ANN-based learning with a high performance-efficiency trade-off in just a few time steps. Our work lights up SNN's potential for 3D point cloud denoising, injecting new perspectives of exploring the deployment on neuromorphic chips while paving the way for developing energy-efficient 3D data acquisition devices.

Paper Structure

This paper contains 25 sections, 10 equations, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Denoise on Point Clouds
  4. Spiking Neural Networks
  5. Spiking Neural Networks on Point Cloud
  6. Methods
  7. Noise-Injected Spiking Neurons
  8. Integrate-and-Fire spiking neurons.
  9. Noise-injected Integrate-and-Fire spiking neurons.
  10. Noise-Injected Spiking Graph Convolution
  11. Graph convolution.
  12. Noise-injected spiking graph convolution (NI-SGC).
  13. Dense NI-SGC block.
  14. Noise-Injected Spiking Graph Convolutional Networks
  15. Experiments and Results
  16. ...and 10 more sections

Figures (5)

  • Figure 1: The probabilistic firing mechanism and calculation process of the NIIF neuron.
  • Figure 2: An illustration of noise-injected spiking graph convolution.
  • Figure 3: An illustration of dense NI-SGC block.
  • Figure 4: Illustration of the architecture and pipeline of the noise-injected spiking point cloud denoise learning network.
  • Figure 5: Visual comparison of denoising methods. Points colored reder are farther away from the ground truth surface.