UAV-CrackX Datasets
收藏资源简介:
这是一个基于无人机的路面裂缝语义分割数据集,采集飞行高度为50米,包含不同放大倍率。公开数据集包含1500张无人机道路图像,其中500张为×4放大倍率(UAV-CrackX4),500张为×8放大倍率(UAV-CrackX8),500张为×16放大倍率(UAV-CrackX16)。同时发布了1200张标注图像(每个放大倍率400张),标注为灰度掩码,其中0表示路面,1表示裂缝。
This is a UAV-based pavement crack semantic segmentation dataset, which is collected at a flight altitude of 50 meters and covers multiple magnification ratios. The publicly available dataset contains 1500 UAV road images, including 500 with ×4 magnification (designated as UAV-CrackX4), 500 with ×8 magnification (UAV-CrackX8), and 500 with ×16 magnification (UAV-CrackX16). Additionally, 1200 annotated images are released, with 400 samples for each magnification type. The annotations are provided as grayscale masks, where pixel value 0 represents the pavement and 1 represents cracks.
UAV-CrackX 数据集概述
数据集基本信息
- 数据集名称:UAV-CrackX Datasets
- 数据类型:基于无人机的路面裂缝语义分割数据集
- 采集高度:50米飞行高度
- 图像总数:1,500张无人机道路图像
数据集组成
放大级别分布
- ×4放大倍率 (UAV-CrackX4):500张图像
- ×8放大倍率 (UAV-CrackX8):500张图像
- ×16放大倍率 (UAV-CrackX16):500张图像
标注信息
- 标注图像数量:1,200张(每个放大级别400张)
- 标注格式:灰度掩码
- 标注类别:
0= 路面1= 裂缝
基准竞赛
- 测试图像:300张(来自不同放大级别)
- 竞赛平台:http://vlp.chd.edu.cn/tasks/4
- 评估内容:参与者可提交分割掩码,在多性能指标上评估模型
相关论文
主要论文
-
IEEE Transactions on Intelligent Transportation Systems:
- 标题:Unmanned Aerial Vehicle (UAV)-Based Pavement Image Stitching Without Occlusion, Crack Semantic Segmentation, and Quantification
- 卷号:25,期号:11,页码:17038–17053,2024年11月
- DOI:https://doi.org/10.1109/TITS.2024.3424525
-
Automation in Construction:
- 标题:Bridging Cross-Domain and Cross-Resolution Gaps for UAV-Based Pavement Crack Segmentation
- 卷号:174,页码:106141,2025年
- DOI:https://doi.org/10.1016/j.autcon.2025.106141
相关研究
-
Advanced Engineering Informatics:
- 标题:GLoU-MiT: Lightweight Global-Local Mamba-Guided U-Mix Transformer for UAV-Based Pavement Crack Segmentation
- 卷号:65,页码:103384,2025年5月
- DOI:https://doi.org/10.1016/j.aei.2025.103384
-
Expert Systems with Applications:
- 标题:DCUFormer: Enhancing Pavement Crack Segmentation in Complex Environments with Dual-Cross/Upsampling Attention
- 卷号:264,页码:125891,2025年3月
- DOI:https://doi.org/10.1016/j.eswa.2024.125891
联系方式
- 联系人邮箱:jhshan@chd.edu.cn




