MagicEyes

MagicEyes是由Magic Leap Inc开发的首个大规模眼睛注视估计数据集，专门为混合现实设备设计。该数据集包含587个不同性别、年龄、种族和眼睛颜色的主题，总计超过800,000张图像，其中80,000张图像具有人工标注的地面实况。数据集的创建旨在推动混合现实设备中眼睛注视估计技术的研究，通过提供多样化的数据支持，帮助开发更准确和鲁棒的学习方法。MagicEyes数据集的应用领域主要集中在提高混合现实设备的交互效率和用户体验，解决在非轴向相机设置下的眼睛注视估计挑战。

MagicEyes is the first large-scale eye gaze estimation dataset developed by Magic Leap Inc., specifically tailored for mixed reality (MR) devices. It contains 587 participants with diverse genders, ages, ethnicities and eye colors, totaling over 800,000 images, 80,000 of which have manually annotated ground truth labels. The dataset was developed to advance eye gaze estimation research for mixed reality devices, providing diversified data support to help build more accurate and robust learning methods. The main applications of the MagicEyes dataset are focused on improving interaction efficiency and user experience of mixed reality devices, and addressing the challenges of eye gaze estimation in non-axial camera setups.