Answers are grounded in this paper and cite their sources.
Drone detection based on radio frequency (RF) signals has become a mainstream method. Spiking Neural Networks (SNNs), known for their superior energy efficiency and temporal dynamic processing capabilities, provide an ideal solution for drone RF signal detection. We propose SpikeRF-YOLO, an low-power and efficient method for detecting drone RF signals. We designe the full-spike SCDSDown module with shortcut branches to mitigate vanishing gradients, improving spectral feature extraction and downsampling efficiency. To better leverage the diverse frequency characteristics of multidrone RF signals, we propose the ESMS module, which merges multi-scale spectral features to enhance detection performance in complex electromagnetic environments. Experimental results show that SpikeRF-YOLO improves mAP@0.5 by 11.4% and 9.3% on the ZK_RF and DroneRFa datasets, while consuming only 15.9% of the power of YOLOv10 and maintaining comparable performance.
No takes yet. Share an insight, caveat, or question.
Liu et al. (2025) studied this question.