AttentionBase: A Facial Landmark-Based Dataset for Visual Attention and Drowsiness Monitoring

AttentionBase is a labeled dataset containing 14,127 samples collected from 200 participants during reading and browsing tasks. It includes facial landmark coordinates, Eye Aspect Ratio (EAR), Mouth Aspect Ratio (MAR), and timestamps, all captured at 20 FPS using MediaPipe, Dlib, and OpenCV. The data is stored in CSV (comma-separated values) format, making it easy to use with a wide range of machine learning and data analysis tools. This dataset is intended for research in visual attention monitoring, drowsiness detection, and human-computer interaction.AttentionBase 是一款标注数据集，共包含14,127条样本，采集自200名参与阅读与网页浏览任务的受试者。该数据集涵盖面部关键点坐标、眼动纵横比（Eye Aspect Ratio, EAR）、嘴部纵横比（Mouth Aspect Ratio, MAR）以及时间戳，所有数据均以20帧每秒的帧率，通过MediaPipe、Dlib与OpenCV工具采集得到。数据集以逗号分隔值（Comma-Separated Values, CSV）格式存储，可轻松适配各类机器学习与数据分析工具。本数据集旨在服务于视觉注意力监测、疲劳检测以及人机交互领域的相关研究。