动态视觉动作识别任务的人眼追踪数据集

本数据集由罗马尼亚科学院数学研究所创建，包含497,107帧视频数据，每帧被16名受试者观看，用于动态视觉动作识别任务。数据集特点包括大规模、视频刺激和任务控制，与自由观看相反。创建过程涉及收集人类在执行视觉动作识别任务时的眼动数据。应用领域主要集中在通过分析人类眼动模式来提高计算机视觉系统的动作识别性能。

This dataset was created by the Mathematical Institute of the Romanian Academy, containing 497,107 video frames. Each frame was viewed by 16 subjects during the dynamic visual action recognition task. The dataset is defined by three core features: large scale, utilization of video stimuli, and task-controlled experimental design, as opposed to free viewing settings. The dataset construction process involved collecting eye-tracking data from human participants while they completed visual action recognition tasks. Its main applications focus on improving the action recognition performance of computer vision systems by analyzing human eye-tracking patterns.