Video2GIF

Video2GIF 数据集包含超过 100,000 对 GIF 及其源视频。 GIF 是从两个流行的 GIF 网站（makeagif.com、gifsoup.com）收集的，相应的源视频是在 2015 年夏季从 YouTube 收集的。提供了 GIF 和视频的 ID 和 URL，以及 GIF 片段的时间对齐到他们的源视频。该数据集将用于评估 GIF 创建和视频突出显示技术。除了 100K GIF 视频对之外，该数据集还包含 357 对 GIF 及其源视频作为测试集。这 357 个视频带有知识共享 CC-BY 许可，允许作者以适当的信用重新分发材料，以使测试集上的结果即使在某些视频不可用时也可重现。

The Video2GIF dataset contains over 100,000 pairs of GIFs and their corresponding source videos. GIFs were collected from two popular GIF platforms: makeagif.com and gifsoup.com, while their source videos were gathered from YouTube in the summer of 2015. The dataset provides IDs and URLs for both the GIFs and their source videos, as well as the temporal alignment information between the GIF clips and their source videos. This dataset is designed to evaluate GIF creation and video highlighting technologies. Apart from the 100,000 GIF-video pairs, the dataset also includes 357 such pairs as a dedicated test set. The 357 source videos in the test set are licensed under the Creative Commons CC-BY license, which permits authors to redistribute the material with proper attribution, ensuring that experimental results on the test set remain reproducible even if some of the videos are no longer available.

搜集汇总

背景与挑战

背景概述

Video2GIF数据集包含超过10万对GIF及其源视频，数据来源于GIF网站和YouTube，提供ID、URL及时间对齐信息，用于评估GIF生成和视频突出显示技术。此外，数据集包含357对采用知识共享许可的测试集，以确保结果可重现。