ddddavidhe/Bilibili-Videos_13k

该数据集包含12587条720p分辨率的视频，每个视频都有唯一的bvid标识、作者名和视频标题。视频文件按照<bvid>.mp4的格式命名，并根据是否从主页爬取或按照up主分类存放在不同目录中。数据集提供了jsonl文件记录视频的bvid、作者名和视频标题的对应关系，以及一个csv文件记录每条视频的时长。由于视频编码问题，使用torchvision读取时可能会出现错误。整个数据集的大小为717GB。The dataset consists of 12,587 720p resolution videos, each with a unique bvid identifier, author name, and video title. The video files are named in the format <bvid>.mp4 and are organized in different directories based on whether they are crawled from the main page or categorized by the up owner. The dataset provides a jsonl file to record the correspondence between video bvid, author name, and video title, as well as a csv file to record the duration of each video. Due to video encoding issues, there may be errors when using torchvision to read. The total size of the dataset is 717GB.

搜集汇总

背景与挑战

背景概述

该数据集包含约12.6k条来自Bilibili平台的720p分辨率视频，总大小约770GB，视频文件以bvid.mp4格式命名，并附带元数据如视频ID、作者、标题和时长。数据按13个类别分类，主要用于视频内容分析和机器学习任务，但需注意视频可能需要重新编码以确保兼容性。