This paper presents an improved version of the YOLOv8s model for road defect detection, addressing the limitations of the original model such as low accuracy, slow detection speed, and high computational complexity. The key contributions include: 1. **EMA Faster Block Structure**: Replacing the Bottleneck structure in the C2f module with a novel EMA Faster Block, which uses partial convolution and multi-scale attention mechanisms to enhance feature representation while reducing computational complexity. 2. **SimSPPF Activation Function**: Replacing the SPPF structure with SimSPPF, which uses the ReLU activation function instead of SiLU, improving detection speed without significant loss in accuracy. 3. **Detect-Dyhead Module**: Introducing a dynamic detection head that dynamically adjusts its focus on different scales and tasks, enhancing precision and recall. The experimental results on the RDD2020 dataset show that the improved model achieves a 5.8% increase in average accuracy (mAP@0.5), a 22.33% reduction in model size, a 23.03% reduction in parameter size, and a 21.68% reduction in computational complexity compared to the original YOLOv8s model. The model also demonstrates superior performance on various datasets, including the United States and Indian datasets, and the GRDDC dataset, showing improved accuracy and reduced false positives and false negatives. The improved YOLOv8s model is suitable for real-time and resource-constrained environments, making it a valuable tool for road maintenance systems.This paper presents an improved version of the YOLOv8s model for road defect detection, addressing the limitations of the original model such as low accuracy, slow detection speed, and high computational complexity. The key contributions include: 1. **EMA Faster Block Structure**: Replacing the Bottleneck structure in the C2f module with a novel EMA Faster Block, which uses partial convolution and multi-scale attention mechanisms to enhance feature representation while reducing computational complexity. 2. **SimSPPF Activation Function**: Replacing the SPPF structure with SimSPPF, which uses the ReLU activation function instead of SiLU, improving detection speed without significant loss in accuracy. 3. **Detect-Dyhead Module**: Introducing a dynamic detection head that dynamically adjusts its focus on different scales and tasks, enhancing precision and recall. The experimental results on the RDD2020 dataset show that the improved model achieves a 5.8% increase in average accuracy (mAP@0.5), a 22.33% reduction in model size, a 23.03% reduction in parameter size, and a 21.68% reduction in computational complexity compared to the original YOLOv8s model. The model also demonstrates superior performance on various datasets, including the United States and Indian datasets, and the GRDDC dataset, showing improved accuracy and reduced false positives and false negatives. The improved YOLOv8s model is suitable for real-time and resource-constrained environments, making it a valuable tool for road maintenance systems.