AI-based Pothole Detection in Adverse Weather

Pothole detection plays a crucial role in ensuring road safety and effective infrastructure maintenance. However, detecting potholes accurately under adverse weather conditions such as rain, fog, snow, and low illumination remains a challenging task due to poor visibility and noise in road images. This project proposes a robust pothole detection system that leverages synthetic images and attention-based object detection techniques to improve detection performance in challenging environments. Synthetic images are generated to simulate diverse weather conditions, enriching the training dataset and reducing dependency on extensive real-world data collection. An attention-based deep learning model is employed to focus on relevant road surface features while suppressing background disturbances caused by weather effects. The proposed approach enhances detection accuracy, robustness, and generalization across varying weather scenarios. Experimental results demonstrate improved performance compared to conventional methods, making the system suitable for real-time road monitoring, intelligent transportation systems, and autonomous driving applications.