MAAD

MAAD数据集由丰田研究院创建，专注于驾驶中的‘注视意识’研究。该数据集包含约24.5小时的注视序列，来自23名参与者在观察驾驶场景视频时的注视数据。此外，数据集还包含了基于参与者扫描路径的第三方注视意识标注。MAAD数据集不仅提供了高精度的注视跟踪，还通过第三方标注探索了参与者的注视和感知到的注视意识。尽管数据收集是通过代理危险意识任务而非真实的驾驶环境进行的，但它为研究驾驶场景中的视觉显著性和意识提供了一个有用的起点。该数据集的应用领域包括提高安全系统中的预警效果，从而增强驾驶员的表现。

The MAAD Dataset was developed by Toyota Research Institute, focusing on research on gaze awareness in driving scenarios. It contains approximately 24.5 hours of gaze sequence data collected from 23 participants while they watched driving scenario videos. Additionally, the dataset includes third-party gaze awareness annotations based on participants' scanpaths. The MAAD Dataset not only provides high-precision gaze tracking, but also allows exploration of participants' gaze and perceived gaze awareness via third-party annotations. While the data was collected through a surrogate hazard awareness task rather than a real driving environment, it serves as a valuable starting point for research on visual saliency and awareness in driving scenarios. Applications of this dataset include improving the warning efficacy of safety systems, thereby enhancing driver performance.