Quick-Sky-Time
收藏资源简介:
Quick-Sky-Time数据集是一个大规模、高分辨率的风景视频数据集,主要包含天空时间流逝场景,如云彩移动的天空和星星移动的星空。该数据集由精心剪辑的216个4K时间流逝视频中的1,167个视频片段组成,每帧分辨率超过1,024×1,024,总帧数达到285,446帧。数据集分为训练、验证和测试集,用于评估不同方法在高质量视频和图像生成任务中的表现,也可作为新的基准。该数据集包含多种挑战性场景,如不同时间段(白天和夜晚)、不同场景对象(山脉、房屋等)、复杂背景以及光线和阴影的变化,适用于多种任务,如高分辨率视频生成、预测、图像生成、纹理生成和图像修复等。
The Quick-Sky-Time dataset is a large-scale, high-resolution landscape video dataset that primarily centers around sky timelapse scenarios, including skies with drifting clouds and starry skies with shifting stars. This dataset consists of 1,167 video clips curated from 216 4K timelapse videos, with each frame having a resolution exceeding 1,024×1,024 and a total of 285,446 frames. The dataset is split into training, validation, and test sets, which are used to evaluate the performance of various methods on high-quality video and image generation tasks, and can also serve as a new benchmark. This dataset encompasses a variety of challenging scenarios, such as different time periods (daytime and nighttime), diverse scene objects (mountains, houses, etc.), complex backgrounds, and changes in lighting and shadows. It is applicable to multiple tasks including high-resolution video generation, prediction, image generation, texture generation, and image inpainting.




