Spiking CenterNet: A Distillation-boosted Spiking Neural Network for Object Detection

TL;DR

This paper introduces Spiking CenterNet, an energy-efficient spiking neural network for object detection that leverages knowledge distillation to improve performance, outperforming previous models on a challenging automotive dataset.

Contribution

It presents the first use of knowledge distillation in spiking neural networks for object detection, combining an SNN adaptation with an efficient decoder for improved accuracy and energy efficiency.

Findings

Outperforms previous SNN-based object detection models on GEN1 dataset.

Uses less than half the energy compared to comparable models.

Knowledge distillation enhances the performance of the spiking neural network.

Abstract

In the era of AI at the edge, self-driving cars, and climate change, the need for energy-efficient, small, embedded AI is growing. Spiking Neural Networks (SNNs) are a promising approach to address this challenge, with their event-driven information flow and sparse activations. We propose Spiking CenterNet for object detection on event data. It combines an SNN CenterNet adaptation with an efficient M2U-Net-based decoder. Our model significantly outperforms comparable previous work on Prophesee's challenging GEN1 Automotive Detection Dataset while using less than half the energy. Distilling the knowledge of a non-spiking teacher into our SNN further increases performance. To the best of our knowledge, our work is the first approach that takes advantage of knowledge distillation in the field of spiking object detection.