360VFI
收藏资源简介:
360VFI数据集由武汉大学计算机科学学院创建,专门用于全方位视频帧插值任务。该数据集包含930个三元组样本,每个样本包含三个视频帧,涵盖多种场景和运动条件。数据集的创建过程中,特别考虑了全方位视频的畸变特性,通过引入畸变先验信息来优化网络性能。360VFI数据集主要应用于增强现实和虚拟现实领域,旨在解决低帧率全方位视频导致的用户体验问题。
The 360VFI dataset, created by the School of Computer Science at Wuhan University, is specifically designed for omnidirectional video frame interpolation tasks. This dataset consists of 930 triplet samples, each containing three video frames and covering diverse scenarios and motion conditions. During the development of this dataset, the distortion characteristics of omnidirectional videos were specially taken into account, and distortion prior information was introduced to optimize the performance of neural networks. The 360VFI dataset is primarily applied in the fields of augmented reality (AR) and virtual reality (VR), with the goal of resolving user experience issues caused by low-frame-rate omnidirectional videos.




