TACoS Multi-Level Corpus

该数据集源自烹饪场景，被用于视频定位和密集视频字幕生成任务。该数据集包含127个视频，平均时长约为4.8分钟，并分别采用10146、4589和4083个视频-查询对进行训练、验证和测试。规模上，该数据集共有127个视频，针对的任务是视频定位和密集视频字幕生成。

This dataset originates from cooking scenarios and is utilized for video grounding and dense video captioning tasks. It comprises 127 videos with an average duration of approximately 4.8 minutes, and uses 10146, 4589 and 4083 video-query pairs for training, validation and testing respectively. In terms of scale, this dataset contains 127 videos, targeting the tasks of video grounding and dense video captioning.

搜集汇总

背景与挑战

背景概述

TACoS Multi-Level Corpus是一个专注于烹饪活动视频的多层次描述数据集，支持生成不同详细程度的连贯多句子描述。该数据集由Max Planck Institute for Informatics发布，仅限科学研究使用，并需遵守特定的许可协议。