TruckV2X
收藏资源简介:
TruckV2X数据集是首个以卡车为中心的多模态和多代理协作感知数据集,旨在解决卡车在自动驾驶中面临的独特感知挑战。该数据集利用LiDAR和摄像头进行多模态感知,并包括拖拉机、拖车、CAV和RSU等多代理协作。数据集提供了64个场景,包括88,396帧LiDAR点云、一百万张相机图像和1.18百万个3D边界框注释。该数据集为开发具有增强遮挡处理能力的协作感知系统奠定了基础,并加速了多代理自动驾驶卡车系统的部署。
The TruckV2X dataset is the first truck-centric multimodal and multi-agent collaborative perception dataset, aiming to address the unique perception challenges faced by trucks in autonomous driving. This dataset leverages LiDAR and cameras for multimodal perception, and covers multi-agent collaboration scenarios including tractors, trailers, CAVs and RSU. The dataset provides 64 scenarios, including 88,396 LiDAR point cloud frames, one million camera images, and 1.18 million 3D bounding box annotations. This dataset lays a foundation for developing collaborative perception systems with enhanced occlusion handling capabilities, and accelerates the deployment of multi-agent autonomous driving truck systems.
TruckV2X数据集概述
基本信息
- 名称: TruckV2X
- 许可证: MIT
- 任务类别: 目标检测
- 标签: 协同感知、自动驾驶卡车、数据集、V2X
- 数据来源: 原创
- 规模: 10K<n<100K
数据集描述
TruckV2X是首个以卡车为中心的大规模协同感知数据集,旨在解决自动驾驶卡车中的独特感知挑战(如大盲区、大型车辆尺寸导致的遮挡以及动态拖车运动)。该数据集具有多模态感知(LiDAR和摄像头)功能,并支持多代理协作,包括拖拉机、拖车、联网和自动驾驶车辆(CAV)以及路边单元(RSU)之间的交互。
使用方法
选项1:通过Hugging Face Hub加载
python from datasets import load_dataset
dataset = load_dataset("XieTenghu1/TruckV2X", trust_remote_code=True)
选项2:克隆+本地解压(推荐)
bash git lfs install git clone https://huggingface.co/datasets/XieTenghu1/TruckV2X cd TruckV2X
解压所有分割文件
for zipfile in train/*.zip; do dirname="${zipfile%.zip}" unzip "$zipfile" -d "$dirname" done
for zipfile in val/*.zip; do dirname="${zipfile%.zip}" unzip "$zipfile" -d "$dirname" done
for zipfile in test/*.zip; do dirname="${zipfile%.zip}" unzip "$zipfile" -d "$dirname" done
数据集结构
TruckV2X/ ├── train/ │ ├── Town1_1/ │ │ ├── cav/ │ │ ├── roadside/ │ │ ├── tractor/ │ │ ├── trailer/ │ ├── Town1_4/ │ ├── ... ├── val/ │ ├── Town3_0/ │ ├── ... ├── test/ │ ├── Town1_0/ │ ├── ...
许可证
- 许可证类型: MIT License
- 许可证链接: https://opensource.org/licenses/MIT




