Fully Spiking Neural Networks with Target Awareness for Energy-Efficient UAV Tracking

2026-03-28

AI Summary

This work addresses the challenge of deploying efficient visual tracking on unmanned aerial vehicles (UAVs) under standard RGB inputs, as existing methods often rely on costly event cameras. To this end, we propose STATrack—the first fully spiking neural network framework tailored for UAV-based RGB visual tracking. STATrack introduces a target-aware mechanism coupled with an adaptive mutual information maximization strategy to strengthen feature correspondence between the template and search regions, thereby enhancing target representation while suppressing background interference. Evaluated on four mainstream UAV tracking benchmarks, the proposed method achieves competitive accuracy and significantly reduces energy consumption, offering a promising pathway toward low-power, edge-deployable tracking systems.

Abstract

Spiking Neural Networks (SNNs), characterized by their event-driven computation and low power consumption, have shown great potential for energy-efficient visual tracking on unmanned aerial vehicles (UAVs). However, existing efficient SNN-based trackers heavily rely on costly event cameras, limiting their deployment on UAVs. To address this limitation, we propose STATrack, an efficient fully spiking neural network framework for UAV visual tracking using RGB inputs only. To the best of our knowledge, this work is the first to investigate spiking neural networks for UAV visual tracking tasks. To mitigate the weakening of target features by background tokens, we propose adaptively maximizing the mutual information between templates and features. Extensive experiments on four widely used UAV tracking benchmarks demonstrate that STATrack achieves competitive tracking performance while maintaining low energy consumption.

Problem

• Spiking Neural Networks
• UAV tracking
• event cameras
• energy-efficient
• visual tracking

Innovation

• Spiking Neural Networks
• UAV Tracking
• RGB Input
• Mutual Information Maximization
• Energy-Efficient