ROAD SAFE-AI:AN INTELLIGENT POTHOLE DETECTION AND ALERT SYSTEM

Road pothole detection is crucial for road maintenance and traffic safety, with deep learning-based computer visionmodels playing a significant role in recent years. However, achieving high detection accuracy remains a challenge due tovariations in road conditions, lighting, and environmental factors. To address these challenges, this study employs theYOLOv8 (You Only Look Once) model for real-time pothole detection. A custom pothole dataset was created,incorporating diverse road conditions, weather variations, and different pothole shapes to improve model generalization.Based on this dataset, we trained and evaluated the YOLOv8 model and compared its performance with previous YOLOversions. The experimental results show that the proposed model achieves an [email protected] of 94.3%, outperformingprevious YOLO models and other traditional detection methods. Additionally, the results indicate that our modelsignificantly reduces false positives while maintaining high detection precision. The findings suggest that YOLOv8 caneffectively identify potholes with improved accuracy, making it suitable for real-time road monitoring and automatedmaintenance systems.