open-sora-pexels-45k

# open-sora-pexels-45k This dataset contains 45k high-quality videos from [Pexels](https://www.pexels.com/). The videos are filtered out from about 400K videos based on different kinds of scores. Pexels.com is a free stock photo and video platform that provides high-quality, royalty-free images and videos for personal and commercial use. All content on Pexels is licensed under the Pexels License, allowing users to download, modify, and use the media without attribution or copyright concerns. There are three `csv` files provided: - `pexels_45k.csv`: contains only path and text, which needs to be processed for training. - `pexels_45k_necessary.csv`: contains necessary information for training. - `pexels_45k_score.csv`: contains score information for each video. The 45k videos are filtered out based on the score. See tech report for more details. For the usage of this dataset, you can visit [Open-Sora](https://github.com/hpcaitech/Open-Sora/tree/upload/v2.0). ## Column interpretation - `path`: path to videos for training - `text`: caption for the video - `num_frames`: number of frames - `height`, `width`, `aspect_ratio`, `resolution`: related to the size of video - `fps`: frame per seconds - `score_aes`: aesthetic scores - `score_vmafmotion`: motion scores, higher means stronger motion in the video - `score_naive_motion`: the probability that the video is a simple zoom-in/out of a picture - `ocr_bbox_ar_max`: the maximum detected text size本数据集包含来自Pexels的4.5万条高质量视频，其从约40万条原始视频中通过多维度评分体系筛选得到。Pexels是一家免费的图库与视频素材平台，可提供高质量、免版税的图片及视频素材，支持个人与商业场景使用。Pexels平台上的所有内容均遵循Pexels许可（Pexels License）协议，用户可在无需署名、无需担忧版权问题的前提下，下载、修改并使用该类素材。如需使用本数据集，请访问：https://github.com/hpcaitech/Open-Sora/

搜集汇总

背景与挑战

背景概述

该数据集是一个包含45K高质量视频的大型视频数据集，体积达539.06GB，采用Apache 2.0开源许可。数据集可通过ModelScope SDK或GIT命令下载，但缺少更详细的内容描述。