Performance Evaluation of YOLOv5 in Adverse Weather Conditions
Fatmanur Ozdemir, Yavuz Selim Bostanci, Müjdat Soytürk · 2023
The technology of object detection is widely utilized in Intelligent Transportation Systems (ITSs) to ensure safe and dependable road experiences. However, adverse weather conditions can pose significant challenges for object detection applications in ITSs. This challenge arises from the traditional machine learning assumption that the training and testing sets have similar data distributions, which may not hold in some real-world situations. In this paper, the YOLOv5 model is trained and tested on the DAWN dataset, which is a diverse object detection dataset concerning weather conditions, to provide an analysis on this domain shift basis. The experiments are also conducted with a transfer learning-based YOLOv5 model to provide comparative analysis.