Multiple Object Eye-Tracking (MOET) dataset

Multiple Object Eye-Tracking (MOET)数据集是由印度理工学院卡拉格普尔分校的研究人员开发的，旨在为注意力解码算法提供一个公开可用的基准。该数据集包含105532个标注帧，这些帧来自于16名参与者在观看14个真实世界视频时的眼动数据，每个视频帧都标注了参与者被分配跟踪的特定对象的类别标签和边界框。MOET数据集的创建过程涉及随机生成目标对象序列，并使用预训练的对象检测器来选择目标对象。该数据集主要用于训练和评估注意力解码模型，特别是在处理复杂视觉刺激和动态视觉环境中的应用。

The Multiple Object Eye-Tracking (MOET) dataset was developed by researchers at the Indian Institute of Technology Kharagpur, aiming to provide a publicly available benchmark for attention decoding algorithms. This dataset contains 105,532 annotated frames sourced from eye-tracking data of 16 participants who watched 14 real-world videos. Each frame is annotated with the category label and bounding box of the specific object that the participant was assigned to track. The development process of the MOET dataset entails randomly generating target object sequences and utilizing pre-trained object detectors to select the target objects. This dataset is primarily utilized for training and evaluating attention decoding models, particularly for applications involving complex visual stimuli and dynamic visual environments.