BIDS

该数据集是一个BIDS数据集，包含8130个视频，每个视频都配有相应的视觉模态（VM）和文本模态（TM）摘要，以及每2秒剪辑的显著性得分注释，这些得分反映了每个剪辑的重要性。此外，该数据集以（视频、TM摘要、VM摘要）的三元组格式构建，确保TM摘要准确捕捉视频的主要内容，而VM摘要则保留显著的视觉内容。这是一个大规模的数据集，包含8130个视频，其任务是进行视频的双模态语义摘要（Bissv）。

This is a large-scale BIDS dataset containing 8,130 videos. Each video is paired with corresponding Visual Modality (VM) and Text Modality (TM) summaries, as well as salience score annotations for every 2-second clip, where the scores reflect the importance of each respective clip. Additionally, the dataset is structured in the triplet format of (video, TM summary, VM summary), ensuring that TM summaries accurately capture the core content of the corresponding videos, while VM summaries retain salient visual content. This dataset is designed for the task of bimodal semantic video summarization (Bissv).