Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces Sequence-SOD, a sequence-aware spiking neural network (SNN) object detector that processes extended event sequences from event cameras, integrating temporal information by preserving membrane potentials across intervals. By accumulating events into short intervals and feeding them sequentially into an SSD-style Spiking DenseNet, the model enhances detection performance while maintaining energy efficiency. The approach significantly improves mean Average Precision (mAP) from 23.38 to 26.88 with data augmentation, demonstrating the effectiveness of leveraging temporal cues in SNNs for object detection tasks.
Sequence-aware training boosts SNN object detection performance by over 3 mAP points, revealing the untapped potential of temporal information in event-based systems.
Event cameras follow a retina-inspired sensing principle, reporting local intensity changes asynchronously with high temporal 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 with event-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.