An Improved YOLOv8 Model for Pavement Crack Detection

This paper introduces an enhanced YOLOv8 model for detecting pavement cracks, utilizing Ghost Convolution and a modified C2f module to reduce computational complexity. The study tackles issues like imbalanced samples and the challenge of accurately detecting small targets. Using a dataset of 2,450 crack images, the improved model showed superior performance compared to the original YOLOv8, particularly in terms of precision, recall, and F1 score, excelling at detecting longitudinal cracks. The enhanced model is optimized for real-time pavement crack detection, offering increased accuracy and robustness across diverse environments.