VLM2Vec/DiDeMo
收藏Hugging Face2025-08-03 更新2025-08-09 收录
下载链接：
https://hf-mirror.com/datasets/VLM2Vec/DiDeMo
下载链接
链接失效反馈官方服务：
资源简介：
DiDeMo数据集包含来自Flickr的10K长格式视频，每个视频都有大约4个按时间顺序注释的短句子组成的长段落地字幕。该数据集用于‘段落到视频’的检索任务，并分为训练集、验证集和测试集。支持的任务类别包括文本到视频、文本检索和视频分类。
The DiDeMo dataset consists of 10K long-form videos from Flickr, each with about 4 short sentences annotated in temporal order to form a long paragraph caption. This dataset is used for paragraph-to-video retrieval tasks and is split into training, validation, and test sets. The supported task categories include text-to-video, text-retrieval, and video-classification.
提供机构：
VLM2Vec
