CoInfra
收藏资源简介:
CoInfra是一个真实世界的大规模合作感知系统和数据集,旨在支持多智能体感知研究,特别是在恶劣天气条件下(雪、雨、冻雨)。它包括14个同步基础设施节点,每个节点配备一个LiDAR和双摄像头,部署在一个共享区域以实现全场景覆盖。数据集包含约195k LiDAR帧、390k摄像头图像、220k标注的3D边界框(涵盖5个类别:汽车、公交车、卡车、自行车、行人)、恶劣天气场景(晴天、雨天、雪天、冻雨)、高清地图以及摄像头-LiDAR和全局对齐的校准数据。
CoInfra is a real-world large-scale cooperation-aware system and dataset designed to support multi-agent perception research, especially under adverse weather conditions including snow, rain, and freezing rain. It includes 14 synchronized infrastructure nodes, each equipped with one LiDAR and dual cameras, deployed in a shared area to achieve full scene coverage. The dataset contains approximately 195k LiDAR frames, 390k camera images, 220k annotated 3D bounding boxes covering 5 categories: car, bus, truck, bicycle, pedestrian, adverse weather scenarios (sunny, rainy, snowy, freezing rain), high-definition maps, as well as calibrated data for camera-LiDAR alignment and global alignment.
CoInfra数据集概述
数据集简介
- 名称:CoInfra (Cooperative Infrastructure Perception System and Dataset)
- 类型:多智能体感知数据集
- 特点:大规模、真实世界、恶劣天气条件下的协同感知
- 应用场景:多节点3D目标检测与跟踪、延迟感知同步、交互式环岛场景运动预测
数据规模与内容
- LiDAR帧数:约195,000帧
- 相机图像:约390,000张(双摄像头)
- 3D标注框:约220,000个
- 标注类别:5类(汽车、公交车、卡车、自行车、行人)
- 天气条件:晴天、雨天、雪天、冻雨
- 附加数据:
- 高清地图(HD map)
- 相机-LiDAR标定数据
- 全局对齐数据
数据集下载
- 示例数据集:Mini Example Dataset
- 完整数据集获取方式:
- 邮件联系:MVS-Lab@uwaterloo.ca
- HuggingFace下载:Download
基准实验结果
| 融合策略 | 使用HD地图 | mAP(所有类别) |
|---|---|---|
| 早期融合 | ✘ | 0.984 |
| 早期融合 | ✔️ | 0.986 |
| 后期融合 | ✘ | 0.931 |
| 后期融合 | ✔️ | 0.952 |
系统构建信息
硬件配置
- LiDAR:Robosense Helios 1615
- 相机:Basler dart daA1920-160uc全局快门RGB相机(双摄像头)
- 计算单元:NVIDIA Jetson Orin NX 16GB
- 供电系统:Renogy Wanderer 10A太阳能充电控制器
- 通信模块:工业级5G调制解调器
软件功能
- Web界面:多节点管理、OTA更新、实时可视化
- ROS传感器节点
- ROS云端系统
引用格式
bibtex @article{ning2025coinfra, title={CoInfra: A Large-Scale Cooperative Infrastructure Perception System and Dataset in Adverse Weather}, author={Ning, Minghao and Yang, Yufeng and Shu, Keqi and Huang, Shucheng and Zhong, Jiaming and Salehi, Maryam and Rahmani, Mahdi and Lu, Yukun and Sun, Chen and Saleh, Aladdin and Hashemi, Ehsan and Khajepour, Amir}, journal={arXiv preprint arXiv:2507.02245}, url={https://arxiv.org/abs/2507.02245}, year={2025} }
许可信息
- 许可证类型:CC BY-NC 4.0




