NVEagle/VideoITG-40K
收藏Hugging Face2025-08-08 更新2025-08-09 收录
下载链接：
https://hf-mirror.com/datasets/NVEagle/VideoITG-40K
下载链接
链接失效反馈官方服务：
资源简介：
VideoITG-40K是一个专门为指令引导的时间定位任务设计的大规模数据集，包含40,000个视频和500,000个注释。该数据集通过自动化注释流程VidThinker构建，整合了视觉和文本理解，适用于视频理解、时间定位、视频问答、多模态AI模型训练和视频检索等应用。
VideoITG-40K is a large-scale dataset specifically designed for instruction-guided temporal grounding tasks, containing 40,000 videos and 500,000 annotations. Constructed through an automated annotation pipeline VidThinker, the dataset integrates visual and textual understanding and is suitable for video understanding, temporal localization, video question answering, multi-modal AI model training, and video retrieval applications.
提供机构：
NVEagle
