PDR
收藏资源简介:
PDR数据集是由中国科学院自动化研究所模式识别国家重点实验室等机构构建的,旨在用于闭环自主驾驶中的决策推理。该数据集包含210k个多样化和高质量的数据样本,通过自动化标注流程捕捉了在Bench2Drive基准测试中的完整决策推理过程,包括场景理解、交通标志识别、关键对象识别和元动作等阶段。所有推理步骤都经过严格的人工验证,确保数据质量。该数据集的发布将为学习结构化和基于因果关系的决策推理提供基础。
The PDR dataset is developed by institutions including the State Key Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, and other relevant entities, with the aim of supporting decision-making and reasoning tasks in closed-loop autonomous driving. It includes 210k diverse, high-quality data samples that capture the full decision-making and reasoning process in the Bench2Drive benchmark via an automated annotation pipeline, covering stages such as scene understanding, traffic sign recognition, key object recognition, and meta-actions. All reasoning steps have undergone strict manual verification to guarantee data quality. The release of this dataset will provide a solid foundation for learning structured and causality-based decision-making reasoning.
ReasonPlan数据集概述
基本信息
- 数据集名称: ReasonPlan
- 研究领域: 自动驾驶
- 主要功能: 统一场景预测与决策推理的闭环自动驾驶系统
技术特点
- 核心能力: 结合场景预测和决策推理的闭环系统
- 技术分类: 计算机视觉(cs.CV)
当前状态
- 代码状态: 正在清理和组织代码
- 开源计划: 将开源全部训练和推理代码
学术引用
- 文献标题: ReasonPlan: Unified Scene Prediction and Decision Reasoning for Closed-loop Autonomous Driving
- 作者: Xueyi Liu等
- 发表年份: 2025
- 文献编号: arXiv:2505.20024




