Cooperative Students: Navigating Unsupervised Domain Adaptation in Nighttime Object Detection

Jicheng Yuan,Anh Le-Tuan,Manfred Hauswirth,Danh Le-Phuoc

2024-05-09

Abstract:Unsupervised Domain Adaptation (UDA) has shown significant advancements in object detection under well-lit conditions; however, its performance degrades notably in low-visibility scenarios, especially at night, posing challenges not only for its adaptability in low signal-to-noise ratio (SNR) conditions but also for the reliability and efficiency of automated vehicles. To address this problem, we propose a \textbf{Co}operative \textbf{S}tudents (\textbf{CoS}) framework that innovatively employs global-local transformations (GLT) and a proxy-based target consistency (PTC) mechanism to capture the spatial consistency in day- and night-time scenarios effectively, and thus bridge the significant domain shift across contexts. Building upon this, we further devise an adaptive IoU-informed thresholding (AIT) module to gradually avoid overlooking potential true positives and enrich the latent information in the target domain. Comprehensive experiments show that CoS essentially enhanced UDA performance in low-visibility conditions and surpasses current state-of-the-art techniques, achieving an increase in mAP of 3.0\%, 1.9\%, and 2.5\% on BDD100K, SHIFT, and ACDC datasets, respectively. Code is available at

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

This paper addresses the performance degradation problem of Unsupervised Domain Adaptation (UDA) in object detection under low-light conditions, especially in nighttime environments. Current methods perform poorly in the adaptation from daytime to nighttime because of the significant differences in lighting, shadows, and contrast in nighttime images, which pose challenges to the reliability and efficiency of applications such as autonomous driving.

To solve this problem, the paper proposes a framework called "Cooperative Students" (CoS). This framework innovatively utilizes the Global-Local Transformations (GLT) module and the Proxy-based Target Consistency (PTC) mechanism. The GLT module enhances daytime images by introducing prior knowledge of nighttime scenes to capture spatial consistency of object features in different scenes. The PTC module iteratively improves the learning quality and pseudo-label quality through the mutual consistency between the teacher network and the proxy student network. In addition, they also develop the Adaptive IoU-informed Thresholding (AIT) strategy to gradually avoid missing potential true positive samples and enrich potential information in the target domain.

Experimental results show that CoS significantly improves the UDA performance under low visibility conditions, achieving improvements of 3.0%, 1.9%, and 2.5% in mAP on the BDD100K, SHIFT, and ACDC datasets, respectively, compared to existing techniques. This demonstrates the effectiveness and advantages of CoS in nighttime object detection.