沥青路面典型病害样本数据集(China Road Damage Detection, CNRDD)包含4319张分辨率为1600*1200的图片，其中训练集3022张，测试集1273张。本数据集中包括了4295张路面病害图片，24张为不包含任何道路损坏的图片。 The CNRDD dataset contains 4319 images with a size of 1600*1200, including 3022 images in the training set and 1273 images in the test set, with a total of 4295 images involved in the training and another 24 images that do not contain any road damage. 道路损坏标注信息存储在一个与图像文件同名的xml文件中，包含的信息包括损坏1-3级别，8种损坏类型和不确定类型，并包含了框坐标信息。 Road damage annotation information is stored in an xml file with the same name as the image file, contains information on damage level 1-3, 8 damage types and uncertain type and box coordinates of road damage. xml文件中道路损坏的框坐标是归一化处理的结果。 The box coordinates of the road damage in the xml file are the result of the normalization process. 损坏标签对应损坏的实际含义如下。 The damage labels correspond to the actual meaning of damage as follows. ------------------------------- Label Meaning ------------------------------- damage_1 Crack damage_3 Longitudinal Crack damage_4 Lateral Crack damage_5 Subsidence damage_6 Rutting damage_8 Pothole damage_9 Looseness damage_11 Strengthening damage_-1 Uncertain ----------------------------------------------------------------------------- XML标签解释： < annotation >: Indicates annotation information starts. < filename >: Indicates the image file corresponding to the annotation information. < object >: Indicates a disease annotation, there can be multiple "< object >" tags in a single annotation file. < damage_* >: When the value is 1 means that the corresponding road damage is of this type. < bndbox >: Coordinates information of the box, which are normalized. < xmin >: x-axis minimum coordinate < ymin >: y-axis minimum coordinate < xmax >: x-axis maximum coordinate < ymax >: y-axis maximum coordinate 使用本数据集请引用相关论文，论文下载： Please cite relevant papers when using this dataset, related papers download： applsci-12-07594.pdf TITS-FSRDD.pdf 论文引用格式(Cite this paper)： [1] Zhang H, Wu Z, Qiu Y, Zhai X, Wang Z, Xu P, Liu Z, Li X, Jiang N. A New Road Damage Detection Baseline with Attention Learning. Applied Sciences. 2022; 12(15):7594. https://doi.org/10.3390/app12157594 [2] Binyi Su, Hua Zhang, Zhaohui Wu, Zhong Zhou. FSRDD: An Efficient Few-Shot Detector for Rare City Road Damage Detection[J]. IEEE Transactions on Intelligent Transportation Systems. 2022, early access, DOI:10.1109/TITS.2022.3208188.