dam crack

The dam crack instance segmentation dataset is constructed from high-resolution UAV images (max 6720×4480 pixels) of Xinfengjiang Dam. Core processing steps include: Preprocessing: Cropping/scaling to 640×640 pixels with contrast enhancement and noise reduction Annotation: 3,258 crack instances manually outlined on 897 images using Roboflow platform Diversity: Contains longitudinal/transverse/network-shaped cracks under varying conditions The dataset focuses on balancing computational efficiency and crack feature preservation, suitable for training robust models for dam safety monitoring.

本大坝裂缝实例分割数据集基于新丰江大坝的高分辨率无人机（UAV）航拍图像构建，单张图像最高分辨率可达6720×4480像素。核心处理流程如下：预处理阶段：将图像裁剪或缩放至640×640像素，并同步执行对比度增强与降噪操作；标注阶段：依托Roboflow平台，在897张图像上人工勾勒出3258个裂缝实例；多样性维度：数据集涵盖不同工况下的纵向裂缝、横向裂缝与网状裂缝。该数据集兼顾计算效率与裂缝特征保留，可用于训练适配大坝安全监测任务的高性能鲁棒模型。