运镜镜头数据集

视频数据集产品为按照影视运镜方法，构图类别等 进行分类标注的视频片段，每个片段两秒以上。视频片段无字幕、水印、台标、logo、边框等画面添加信息。此片段仅供AI大模型数据训练场景使用，可提供视频片段1万条，其中包括影视作品成片截取的片段，和公司自行拍摄的视频片段。应用场景为Ai大模型训练，包括对人物形象、服饰装扮、动作、语言、口型、音乐、音效等针对性训练。This video dataset product consists of video clips classified and annotated based on cinematic shooting techniques, composition categories and other relevant standards. Each clip lasts more than 2 seconds, and contains no added visual elements such as subtitles, watermarks, TV channel logos, brand logos or frame borders. This dataset is exclusively intended for use in AI large language model (LLM) data training scenarios. A total of 10,000 video clips will be provided, including clips extracted from finished film and television works and self-shot videos produced by the company. Its application scenarios focus on AI LLM training, including targeted training on human figures, clothing and styling, movements, speech, lip movements, music, sound effects and other related aspects.

搜集汇总

背景与挑战

背景概述

该数据集包含1万条2秒以上的无干扰视频片段，按影视运镜和构图分类标注，含影视截取和自制内容，专用于AI大模型对人物形象、动作、音效等要素的训练。