中国移动通识视频数据集

本数据集为高质量通识类视频描述数据集，涵盖各类互联网创作者发布的视频作品，同步配套创作者撰写的视频文本描述与标签列表，视频内容兼具美感与场景覆盖度。内容覆盖自然、旅行、休闲娱乐、体育锻炼、城市、商业、食物、教育科学、家庭生活、医疗健康、建筑空间、工业、活动、表演、农业、科技产品、交通运输、动物、美容时尚、影视、艺术、购物、人物、手工艺、宗教、家居用品、军事、灾害、广播主持、工具等全品类日常通识场景，样本具备强场景属性、视频文本匹配属性与全品类覆盖特性，适合用于多模态大模型训练、视频内容理解、视频分类、智能标签生成、视频内容检索、智能推荐系统构建及视频内容合规审核。

This is a high-quality general knowledge video description dataset, covering video works released by various online creators, and paired with the video text descriptions and tag lists written by the creators themselves. The video content balances aesthetic value and wide scene coverage. Its content spans all categories of daily general knowledge scenarios, including nature, travel, leisure and entertainment, physical exercise, urban life, commerce, food, educational science, family life, medical and health care, architectural spaces, industry, events, performances, agriculture, technology products, transportation, animals, beauty and fashion, film and television, art, shopping, figures, handicrafts, religion, household goods, military affairs, disasters, radio hosting, tools, etc. The dataset samples possess strong scene relevance, excellent video-text alignment, and full-category coverage characteristics. It is applicable to multi-modal large language model training, video content understanding, video classification, intelligent tag generation, video content retrieval, intelligent recommendation system development, and video content compliance auditing.

搜集汇总

背景与挑战

背景概述

中国移动通识视频数据集是一个高质量的通识类视频描述数据集，包含互联网创作者发布的视频作品及其配套的文本描述与标签列表，覆盖自然、旅行、休闲娱乐等全品类日常场景。该数据集适用于多模态大模型训练、视频内容理解、分类、智能标签生成、检索、推荐系统构建及合规审核等多种应用场景。