PRB-FPN+: Video Analytics for Enforcing Motorcycle Helmet Laws

Bor-Shiun Wang, Ping-Yang Chen, Yi-Kuan Hsieh, Jun-Wei Hsieh, Ming-Ching Chang, JiaXin He, Shin-You Teng, HaoYuan Yue, Yu-Chee Tseng; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023, pp. 5477-5485

Road safety is of utmost importance, and helmet compliance for motorcyclists plays a crucial role in mitigating severe injuries and fatalities. Monitoring and enforcing helmet usage remains a challenge due to the sheer volume of motorcyclists and limited resources for enforcement. Real-time object detection technology offers a promising solution for monitoring helmet usage by effectively identifying motorcyclists and their adherence to helmet rules. However, accurately discerning helmeted motorcyclists and determining driver and passenger positions in complex real-world scenarios remains a challenge. In this paper, we present a novel two-step approach to address these challenges. First, we introduce the PRB-FPN+, a state-of-the-art detector that excels in object localization. We also explore the benefits of deep supervision by incorporating auxiliary heads within the network, leading to enhanced performance of our deep learning architectures. Second, we employ an advanced tracker to refine the tracking of drivers and passengers. Comprehensive experimental results demonstrate that our PRB-FPN+ outperforms current state-of-the-art methods, approaching the highest performance levels. Our proposed system achieved a rank of 8 on the Leaderboard in the AI City Challenge 2023 Track 5. This streamlined approach aims to provide a more reliable and accurate solution for monitoring and enforcing helmet rules among motorcyclists in challenging real-world environments.