DR(eye)VE

该数据集名为DR(eye)VE，专注于捕捉典型的驾驶场景，尤其关注在驾驶过程中LRP热力图与人类视线之间的相关性。该数据集强调在驾驶过程中观察到的与任务相关的刺激，并允许比较不同的视线预测技术。其规模覆盖了典型的驾驶情境，任务包括驾驶行为预测和视线预测。

This dataset, named DR(eye)VE, focuses on capturing typical driving scenarios, with particular attention paid to the correlation between LRP heatmaps and human gaze during driving. It emphasizes task-relevant stimuli observed during driving, and allows for comparisons between different gaze prediction technologies. It covers a range of typical driving contexts, and its associated tasks include driving behavior prediction and gaze prediction.

搜集汇总

背景与挑战

背景概述

DR(eye)VE数据集是一个专注于驾驶员注意力研究的视频数据集，包含74段5分钟的视频序列，标注了驾驶员的注视点和显著性图。该数据集具有高度多样性，涵盖8名驾驶员在不同场景、天气和时间条件下的驾驶行为，并特别包含了注意力集中和分散的子集，适用于自动驾驶和辅助驾驶相关研究。