SCOPE (Stereoscopic COntent Preference Evaluation)
收藏资源简介:
SCOPE数据集由苹果公司、魏茨曼科学研究所和麻省理工学院联合创建,旨在评估立体图像的质量和用户体验。该数据集包含2400个样本,涵盖了多种常见的感知扭曲和生成方法,数据来源包括物理捕捉的立体图像和通过生成方法处理的单目图像。数据集通过VR头显收集用户的偏好标注,用于训练立体体验质量评估模型iSQoE。该数据集的应用领域主要集中在虚拟现实和立体图像生成技术中,旨在解决立体图像质量评估中的数据瓶颈问题,提升用户体验。
The SCOPE dataset was co-developed by Apple Inc., the Weizmann Institute of Science, and the Massachusetts Institute of Technology (MIT), with the goal of evaluating the quality and user experience of stereoscopic images. Comprising 2400 samples, the dataset covers a variety of common perceptual distortions and generation methods, with data sources including physically captured stereoscopic images and monocular images processed via generative approaches. User preference annotations were collected using VR headsets, and the dataset is utilized to train the stereoscopic experience quality assessment model iSQoE. Its primary application scenarios focus on virtual reality and stereoscopic image generation technologies, aiming to address the data bottleneck issue in stereoscopic image quality assessment and enhance user experience.




