ekazakos/iGround
收藏Hugging Face2025-11-09 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/ekazakos/iGround
下载链接
链接失效反馈官方服务：
资源简介：
iGround数据集是一个手动注释的数据集，用于视频到文本的任务。该数据集包含英语语言的数据，并包括文本生成、视频字幕和视频定位等标签。数据集的大小在1,000到10,000之间。数据集有三种配置：处理过的数据、原始数据和键值数据，每种配置都包括训练集、验证集和测试集。该数据集在论文《大规模预训练用于视频字幕生成》中介绍。
The iGround dataset is a manually annotated dataset used for video-text-to-text tasks. It contains data in English and includes tags such as text generation, video captioning, and video grounding. The dataset size is between 1,000 and 10,000. There are three configurations of data: processed, raw, and keys, each including train, validation, and test splits. The dataset is introduced in the paper Large-scale Pre-training for Grounded Video Caption Generation.
提供机构：
ekazakos
