RZDG (Roadwork Zone Detection and Geo-localization)
收藏资源简介：
RZDG数据集由因戈尔施塔特应用技术大学等机构联合构建，旨在解决自动驾驶中临时施工区域检测与地理定位的难题。该数据集包含真实世界与仿真数据，涵盖1357帧点云与859帧图像，总计标注了12480个三维边界框和43个对象的全局坐标，数据来源于德国道路施工场景及CARLA模拟环境。数据集通过多传感器同步采集与精细标注流程构建，支持图像语义分割、三维目标检测及对象地理定位等任务，主要应用于自动驾驶系统的高清地图更新与动态环境感知，以提升行车安全与导航效率。
The RZDG dataset was jointly constructed by the University of Applied Sciences Ingolstadt and other institutions, aiming to address the challenges of temporary construction zone detection and geolocation in autonomous driving. This dataset includes both real-world and simulated data, covering 1357 frames of point cloud data and 859 frames of images, with a total of 12480 annotated 3D bounding boxes and global coordinates for 43 objects. The data is sourced from real road construction scenarios in Germany and the CARLA simulation environment. The dataset is developed through multi-sensor synchronous collection and meticulous annotation processes, supporting tasks including image semantic segmentation, 3D object detection and object geolocation. It is primarily applied to high-definition map updating and dynamic environment perception for autonomous driving systems to improve driving safety and navigation efficiency.
RZDG 数据集概述
RZDG（Roadwork Zone Detection and Geo-localization）是一个多模态数据集，专为支持道路作业区感知、检测、跟踪和地理定位研究而设计。
数据构成
RZDG 数据集包含两大部分：
- RZDG-real（真实数据）：包含 3643 个路标和 8837 个路障，涵盖 28 个不同的道路作业区场景，共 1357 帧。
- RZDG-sim（仿真数据）：包含 419 个路标和 254 个路障，涵盖 8 个模拟城镇中的 203 个不同道路作业区场景，共 8591 帧。
数据格式
数据集采用 KITTI 格式组织，目录结构如下：
RWDG_dataset ├── ImageSets ├── kitti_gt_database ├── testing │ ├── calib │ ├── image_0 │ └── lidar └── training ├── calib ├── image_0 ├── label_0 └── lidar
支持任务
- 图像语义分割
- 3D 目标检测与跟踪
- 道路作业区地理定位
下载地址
检测与跟踪框架
- 3D 目标检测：采用 MMDetection3D
- 3D 目标跟踪：采用 AB3DMOT
为适配 RZDG 数据集，提供了自定义补丁文件，位于 patches/ 目录下，用于修改上述框架以支持 RZDG 的数据格式、目标类别和评估设置。
论文引用
bibtex @inproceedings{yan2025framework, title={Framework and Multi-modal Dataset for Roadwork Zone Detection and Geo-localization}, author={Yan, Zhiran and Xin, Yutong and Shenoi, S Shyam and Song, Rui and Elger, Gordon}, booktitle={2025 IEEE Intelligent Vehicles Symposium (IV)}, pages={1758--1765}, year={2025}, organization={IEEE} }
