Tencent-MVSE

腾讯MVSE是用于多模态视频相似性评估任务的大规模基准数据集。腾讯MVSE的特点包括
总共1,135,705个视频，包括100万视频用于无人监督的预培训63,613视频用于培训，以及63,960视频用于评估。
每个视频的丰富元数据，包括标题、ASR文本、类别和标签。
我们注释67,854视频对用于训练和67,887视频对用于评估。
328类别和64,903标签都是人工注释的。

Tencent MVSE is a large-scale benchmark dataset for multimodal video similarity evaluation tasks. Its core characteristics are as follows: 1. It contains a total of 1,135,705 videos, including 1 million videos for unsupervised pre-training, 63,613 videos for model training, and 63,960 videos for evaluation. 2. Each video is equipped with rich metadata, including title, ASR text, category and tags. 3. We annotated 67,854 video pairs for training and 67,887 video pairs for evaluation. 4. A total of 328 categories and 64,903 tags were manually annotated.

搜集汇总

背景与挑战

背景概述

Tencent-MVSE是一个公开数据集，由OpenDataLab发布，大小为811.0B，在CVPR会议中有所关联，但缺乏详细的介绍内容，无法进一步提炼其具体信息或特点。