Weather-Augmented Traffic Sign Classification with YOLOv3: A Custom Dataset Approach
Suresh Kumar Samarla, Antharaju K Chakravarthy, Phanikumar, Sameena Begum · 2025
Advanced driver-assistance systems (ADAS) and autonomous driving systems depend on the identification of traffic signs, however ,unfavorable weather patterns such as rain, dust, snow, and sunlight pose substantial challenges for accurate detection. This paper presents a novel custom dataset simulating these extreme climatic conditions and employs a YOLOv3-based classification model to detect traffic signs. The model obtained an accuracy of 92.74%, with a recall of 92.74%, precision of 93.55%, and F1-score of 92.62%. These findings demonstrate the strength of the model in handling challenging real-world conditions and present an effective approach to improving traffic sign detection in adverse climate scenarios.