Multi-Object Tracking in the Dark

• DOI: 10.48550/arxiv.2405.06600
• Publication Date: 2024-05-10
• Authors: Xinzhe Wang; Kang Ma; Qiankun Liu; Yunhao Zou; Ying Fu
• Keywords: Tracking (education)

Low-light scenes are prevalent in real-world applications (e.g. autonomous driving and surveillance at night). Recently, multi-object tracking various practical use cases have received much attention, but dark is rarely considered. In this paper, we focus on scenes. To address the lack of datasets, first build a Multi-Object Tracking (LMOT) dataset. LMOT provides well-aligned low-light video pairs captured by our dual-camera system, high-quality annotations for all videos. Then, propose method, termed as LTrack. We introduce adaptive low-pass downsample module to enhance low-frequency components images outside sensor noises. The degradation suppression learning strategy enables model learn invariant information under noise disturbance image quality degradation. These improve robustness conducted comprehensive analysis dataset proposed Experimental results demonstrate superiority method its competitiveness real night Dataset Code: https: //github.com/ying-fu/LMOT