Pavement Surface Distress Point Cloud Dataset: Cracks, Potholes, and Upheavals

This study constructed a point cloud dataset focusing on pavement distresses such as cracks, potholes, and upheavals, collected from national and provincial highways. The data were acquired onsite using a line-structured light sensor combined with an encoder, resulting in a total of 533 point cloud samples of various distresses. After preprocessing, the collected data were annotated using CloudCompare, a professional point cloud processing software. To ensure balanced distribution and enhance the generalization capability of the network model during training, the dataset was divided into training and validation sets in a ratio of 8:2 according to the number of point clouds in each category. These subsets are designated as Area_1 and Area_2, respectively. The dataset is formatted following the S3DIS standard, with text files stored in .txt format.The text contains six columns of data, with the first three columns representing three-dimensional coordinates and the last three columns representing corresponding normal vector coordinates.