S-VideoXum
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/IDT-ITI/SD-Vsum
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是专为脚本驱动的视频摘要而扩展的VideoXum数据集版本，包含了视频样本及对应的自然语言描述的真实摘要。此外，该数据集还包含了使用LLaVA-NeXT-7B模型生成的自然语言摘要，其中训练集分布有6,782个样本，验证集有3,419个，测试集有1,707个。整个数据集规模达到了11,908个视频，其任务定位于脚本驱动的视频摘要。
This dataset is an extended version of the VideoXum dataset specifically designed for script-driven video summarization. It includes video samples paired with their corresponding ground-truth natural language abstracts. Additionally, the dataset also contains natural language abstracts generated using the LLaVA-NeXT-7B model. The training set consists of 6,782 samples, the validation set has 3,419 samples, and the test set contains 1,707 samples. The entire dataset comprises 11,908 video samples in total, with its task oriented towards script-driven video summarization.
