CoCAtt

CoCAtt数据集是由伊利诺伊大学厄巴纳-香槟分校的研究团队创建，专注于驾驶员注意力预测，特别是在自动驾驶和手动驾驶模式下的应用。该数据集包含约11.88小时的驾驶员注意力数据，涵盖了不同自主级别的驾驶场景，使用不同分辨率的眼睛跟踪设备同时收集数据。数据集的创建旨在解决现有模型在驾驶员注意力预测中忽视驾驶员分心状态和意图的问题，通过提供每帧的驾驶员状态注释，包括分心状态和意图，来增强驾驶员注意力模型的性能。CoCAtt数据集的应用领域主要集中在提高自动驾驶系统的安全性和效率，通过更准确地预测驾驶员的注意力行为，以预防高风险事件如碰撞和伤亡。

Created by a research team at the University of Illinois Urbana-Champaign, the CoCAtt Dataset focuses on driver attention prediction, with particular emphasis on applications in both autonomous and manual driving modes. This dataset encompasses approximately 11.88 hours of driver attention data, covering driving scenarios across different levels of autonomy, and is concurrently collected using eye-tracking devices with varying resolutions. The development of this dataset aims to resolve the critical issue that existing driver attention prediction models fail to account for drivers' distraction states and intentions. By supplying per-frame driver state annotations that include distraction status and driving intentions, it seeks to boost the performance of driver attention models. The main application scenarios of the CoCAtt Dataset center on enhancing the safety and efficiency of autonomous driving systems. By more precisely predicting drivers' attention behaviors, the dataset enables the prevention of high-risk events such as collisions, fatalities and injuries.