Accurate and timely detection of pavement distress plays a crucial role in ensuring road safety, effective maintenance planning, and thorough evaluation of pavement conditions. Efficient distress detection and classification methods are urgently needed to optimize maintenance costs and ensure long-term road safety. AI approaches have shown promise in this area in recent years. Object detection algorithms are playing an important role in this topic. A practical, fast, and accurate multi-type pavement distress detection model is needed. To address these issues, we propose a novel YOLOv8-attention pavement distress detector. Our study established a comprehensive dataset containing seven pavement distress types. The dataset comprises 3914 images, annotated with 11424 bounding boxes, providing detailed information on the type and location of distress instances. Based on YOLOv8n, we added different attention modules, including Multi-Head Self-Attention (MHSA) module and the Selective kernel attention (SKA) module. On the same pavement distress dataset we establish, the AP result of YOLOv8-MHSA is 62.6%. The AP result of YOLOv8-SKA is 64.3%. The AP results are 1.63% and 3.32% higher than the original YOLOv8n model in the same 100 training epochs. It was observed that most of the misclassifications occur between the background class and the different distress classes. Consequently, future research will consider dividing the process into a two-step classification algorithm for distress detection followed by a distress-type classification stage.