This dataset presents pavement distress data collected using high-altitude Unmanned Aerial Vehicles (UAVs) over road networks in Shanxi, China. The data collection involved capturing aerial images of road pavements with UAVs flying at high altitudes to efficiently cover large areas. A total of 11,696 high-resolution road pavement images were acquired and annotated with detailed distress information: 12,365 line annotations indicating linear cracks, 8239 block annotations marking block cracks, and 1412 pit annotations identifying potholes. Named HighRPD, this extensive dataset addresses the scarcity of publicly available UAV-based road pavement distress datasets, which are currently limited in data volume. HighRPD offers a substantial number of samples compared to existing public datasets, providing a valuable resource for developing and benchmarking pavement distress detection algorithms. Additionally, the dataset offers data scientists and machine learning engineers a rich repository of road surface data, facilitating the development and training of models for image recognition, pavement condition classification, and object detection. Consequently, HighRPD supports applied research in areas such as transportation and urban planning.
Keywords: Computer vision; Crack; Object detection; Pavement distress; Pothole; Unmanned aerial vehicle (UAV).
© 2025 The Author(s).