Weakly Supervised Road Mark Detection for Advanced Driver Assistance Systems in Consumer Automotive Technology

Road mark detection is a critical component of Advanced Driver Assistance Systems (ADAS), ensuring accurate lane-keeping, navigation, and driver safety. However, the high cost of acquiring fully annotated datasets for supervised learning poses a significant challenge for real-world applications. This paper proposes a novel weakly supervised learning method for road mark detection tailored to consumer automotive technology. Leveraging weak annotations, the method efficiently learns to detect and classify road marks with minimal human intervention. The approach was evaluated on the publicly available CeyMo dataset, achieving a classification accuracy of 99.3% and a localization accuracy of 71.5%, demonstrating its effectiveness and practicality. The proposed method reduces the dependency on fully annotated data while maintaining competitive performance, making it a scalable solution for consumer-level ADAS applications. This work contributes to advancing cost-effective, reliable, and scalable ADAS technologies, paving the way for broader adoption in intelligent vehicles.