Multi-Scene Road Defect Dataset

This study utilizes an independently collected road defect dataset, comprising images from in-vehicle cameras recording traffic on highways, rural streets, and town roads across multiple areas of Jiangxi Province, China. The dataset includes 4,374 original images, from which 3,208 defect samples were identified through data screening. Initially, these images were manually labeled into six categories such as transverse cracks, longitudinal cracks, cracks, blurred markings, potholes and splice cracks. Due to privacy and intellectual property considerations. Only a subset of the custom datasets is provided for demonstration purposes in the `datasets/partial_dataset` directory. The public datasets in `datasets/public_dataset` can be downloaded from their respective sources.

本研究采用自主采集的道路缺陷数据集，其数据来源于中国江西省多个区域的车载摄像头所录制的高速公路、乡村街道及城镇道路的交通影像。该数据集初始包含4374张原始图像，经数据筛选后共识别出3208张缺陷样本。最初，研究人员通过人工标注将这些图像划分为六大类别，分别为横向裂缝、纵向裂缝、裂缝、标识模糊、坑洞及拼接裂缝。出于隐私保护与知识产权考量，本研究仅在`datasets/partial_dataset`目录中提供部分自定义数据集用于演示。`datasets/public_dataset`目录下的公开数据集，可从其各自的来源渠道下载。