Research on Road Damage Detection Based on an Improved YOLOv8 Model

This paper presents the YOLOv8 model for road damage detection, highlighting performance enhancements. YOLOv8 features a novel backbone network, an anchor-free detection head, and new loss functions, ensuring efficient operation across various hardware platforms. A self-attention mechanism is integrated to enhance classification accuracy and robustness. Experimental comparisons indicate that while the attention-augmented YOLOv8 improves precision and recall, the modified version achieves superior overall performance. Analyses using confusion matrices and precision-recall curves demonstrate the enhanced YOLOv8's greater accuracy and stability with complex backgrounds and diverse damage types. The results confirm its superiority in practical applications, showcasing strong performance in road damage detection.