SUT-Crack

The SUT-Crack dataset contains high-quality images of asphalt pavement cracks, specifically curated for crack detection using deep learning techniques like classification, object detection, and segmentation. Careful consideration was given during dataset creation to encompass various crack detection challenges, such as oil stains, shadows, and different lighting conditions. The images were captured from a fixed height of 672 mm above the pavement surface, facilitating calibration for real-world crack length measurements. Notably, the dataset also includes geotags, providing precise latitude and longitude coordinates for each image.

SUT-Crack数据集（SUT-Crack dataset）包含高质量的沥青路面裂缝图像，专为采用分类（classification）、目标检测（object detection）、分割（segmentation）等深度学习技术的裂缝检测任务定制。数据集构建过程中充分考量了裂缝检测场景中的各类挑战，涵盖油污、阴影与不同光照条件等干扰因素。所有图像均以路面上方672毫米的固定高度拍摄，便于后续开展实际裂缝长度的校准测量。值得注意的是，该数据集还附带地理标签（geotags），可为每张图像提供精确的经纬度坐标。