Image-PointClouds Multimodal Crack Segmentation Dataset

To construct a multimodal dataset of images and point clouds, the Intel RealSense L515 LiDAR camera was used for multi-scene acquisition. The sensor can synchronously capture image and point cloud data. This dataset covers typical crack scenarios under diverse complex backgrounds, including concrete cracks on building surfaces, underground coal mine roadway surrounding rock cracks, and pavement cracks, encompassing variations in material, illumination, surface contamination, and occlusion, while ensuring representative crack orientations and morphologies to reflect the realism of model testing in actual engineering environments.