Bridge Crack Library

The bridge crack library (BCL) was an image dataset established to support the deep learning-based crack detection research. This dataset was based on 1180 raw images collected from inspection for in-service bridges. It contains 11000 pixel-wise labeled images with 256×256 resolutions. The dataset could be used to the training of deep neural networks for semantic segmentation of crack images.桥梁裂缝库（Bridge Crack Library, BCL）是为支撑基于深度学习的裂缝检测研究而构建的图像数据集。该数据集基于从在役桥梁巡检中采集的1180张原始图像，包含11000张分辨率为256×256的逐像素标注图像。本数据集可用于训练面向裂缝图像语义分割的深度神经网络。