Sequence-SOD: Bio-inspired Sequence-aware Spiking ObjectDetection for Event Cameras

2026-07-29

AI Summary

This work addresses the limitation of conventional spiking neural networks (SNNs) in event-based object detection, which typically process isolated temporal windows and reset internal states, thereby failing to exploit the rich temporal dynamics of continuous event streams. To overcome this, we propose a sequence-aware SNN framework that, for the first time, introduces sequence-aware training to this domain by maintaining membrane potentials across long event sequences. Integrated within an SSD-style Spiking DenseNet architecture, our approach leverages event accumulation, timestep discretization, and data augmentation to enable evolving neural states for continuous detection. Evaluated on the Gen1 dataset, the method improves mean average precision (mAP) from 23.38 to 25.30 without augmentation and to 26.88 with augmentation, while achieving a theoretical inference rate of 40 Hz—enhancing temporal modeling without compromising the inherent energy efficiency of SNNs.

Abstract

Event cameras follow a retina-inspired sensing principle, reporting local intensity changes asynchronously with hightemporal resolution and a wide dynamic range. Spiking Neural Networks (SNNs) complement these sparse event streams through brain-inspired dynamics, using sparse spikes and leaky membrane potentials to integrate information over time. However, many SNN object detectors process isolated event intervals with a single label and reset the network state after each prediction, thereby underusing temporal information in continuous event streams. We introduce Sequence-SOD, a sequence-aware SNN object detector that processes extended event sequences containing labels at multiple time points. Events are accumulated into short intervals, discretized into temporal steps, and fed sequentially to an SSD-style Spiking DenseNet while preserving membrane potentials across intervals within a sequence, so that detection is driven by an evolving neural state instead of independently reset input windows. On the Gen1 Automotive Detection Dataset, sequence-aware training improves mAP from 23.38 for single-interval training to 25.30 without augmentation and to 26.88 withevent-data augmentation. The model achieves a theoretical prediction frequency of 40 Hz. Training and evaluating SNN object detectors on extended event sequences improves their ability to exploit temporal cues while preserving the energy-efficiency benefits of sparse spiking computation. The results highlight sequence-aware training as a complementary direction to architectural improvements for event-based SNN detection.

Problem

• event cameras
• Spiking Neural Networks
• object detection
• temporal information
• sequence-aware

Innovation

• sequence-aware
• spiking neural networks
• event cameras
• temporal integration
• object detection