本田美国研究院驾驶数据集 (HDD)
收藏资源简介:
本田美国研究院驾驶数据集(HDD)是由波士顿大学和本田美国研究院合作创建的,旨在研究真实环境中的驾驶员行为和因果推理。该数据集包含104小时的旧金山湾区真实人类驾驶数据,通过装备有多种传感器的车辆收集。数据集分为137个会话,每个会话代表一个驾驶员执行的导航任务。HDD通过引入4层注释方案来描述驾驶员行为,包括目标导向行动、刺激驱动行动、原因和注意力,旨在解决智能交通系统中驾驶员行为理解的问题。
The Honda Research Institute USA Driving Dataset (HDD) was co-developed by Boston University and Honda Research Institute USA to study driver behavior and causal reasoning in real-world scenarios. This dataset encompasses 104 hours of real human driving data collected from the San Francisco Bay Area using instrumented vehicles equipped with multiple sensors. It is divided into 137 sessions, each corresponding to a navigation task completed by an individual driver. HDD adopts a four-layer annotation schema to characterize driver behaviors, including goal-directed actions, stimulus-driven actions, causal antecedents, and attentional states, aiming to address the challenges of driver behavior understanding in intelligent transportation systems.

- 1Toward Driving Scene Understanding: A Dataset for Learning Driver Behavior and Causal Reasoning波士顿大学, 本田美国研究院 · 2018年



