In the current era of rapid urbanization, the development of urban road traffic infrastructure is proceeding at an unprecedented rate, significantly enhancing travel convenience. However, road maintenance issues, particularly potholes, have become increasingly troublesome. Potholes can disrupt the smooth and safe passage of vehicles and pose hazards to pedestrians. Consequently, the issue of potholes has attracted more attention. Traditional detection methods predominantly involve manual inspections, which are laborious, inefficient, prone to subjective biases, and can hinder traffic flow. To address existing challenges in pothole detection, we propose using the YOLOv8 model, augmented with a multilayer attention mechanism to refine the identification of road surface defects. By incorporating a pre-trained YOLOv8 model with this attention mechanism, the emphasis on detail is enhanced by increasing the weight of features across different layers. This strategy bolsters the model’s ability to represent features and detect accurately. Compared to the standard YOLOv8 model, the enhanced version significantly reduces the training time for pothole detection while maintaining an accuracy rate of approximately 85%. The Optimization Algorithm for Road Pothole Recognition, which integrates a Multi-Layer Attention Mechanism in YOLOv8, facilitates more efficient and precise road defect detection in real-world applications, underscoring its substantial practical value.