HiDream-ai/ReCo-Data

ReCo-Data是一个大规模、高质量的视频编辑数据集，包含超过50万个指令-视频对。该数据集主要用于视频编辑任务，包括替换、风格化、移除和添加等任务。数据集的收集流程包括六个主要阶段：原始数据预处理、对象分割、指令生成、条件对构建、视频合成和视频过滤与重新标注。数据集的结构按照任务类型组织，每个任务类型下有源视频和目标视频的文件夹。数据集的使用需要遵循CC BY-NC-SA 4.0许可证，仅用于学术研究目的。

ReCo-Data is a large-scale, high-quality video editing dataset comprising 500K+ instruction-video pairs. This dataset is primarily used for video editing tasks, including replace, style, remove, and add tasks. The data collection pipeline consists of six primary stages: raw data pre-processing, object segmentation, instruction generation, condition pair construction, video synthesis, and video filtering and re-captioning. The dataset is organized by task type, with each task type containing folders for source videos and target videos. The dataset is licensed under CC BY-NC-SA 4.0 and is intended for academic research purposes only.