SAM-DA: UAV Tracks Anything at Night with SAM-Powered Domain Adaptation
核心概念
SAM-DA proposes a novel framework for real-time nighttime UAV tracking using SAM-powered domain adaptation, enhancing training sample quality and reducing the reliance on raw images.
摘要
I. Introduction
Object tracking crucial for UAV applications.
SOTA trackers struggle in low-light conditions.
Domain adaptation introduced for nighttime UAV tracking.
II. Proposed Method
SAM-powered DA framework introduced.
SAM-powered target domain training sample swelling designed.
Experiments validate robustness of SAM-DA for nighttime UAV tracking.
III. Experiments
Implementation details and data used.
Evaluation on DarkTrack2021 and NUT-L benchmarks.
Analysis on training sample swelling and fewer-better training.
IV. Conclusion
Introduction of SAM into day-night domain adaptation enhances tracker performance.
V. References
"An innovative SAM-powered target domain training sample swelling is designed to determine enormous high-quality target domain training samples from every single challenging nighttime image."
"Comprehensive experiments on extensive nighttime videos prove the robustness and domain adaptability of SAM-DA for nighttime UAV tracking."