MBZUAI/VCG-plus_112K
收藏Hugging Face2024-06-17 更新2024-06-12 收录
下载链接:
https://hf-mirror.com/datasets/MBZUAI/VCG-plus_112K
下载链接
链接失效反馈官方服务:
资源简介:
VCG+ 112K数据集是为了改进VideoInstruct100K数据集的标注过程而开发的。通过改进关键帧提取、利用最先进的大型多模态模型(LMMs)进行详细描述以及优化指令生成策略,该数据集提高了指令调校对的准确性和质量。
The VCG+112K dataset is developed through an improved semi-automatic annotation pipeline aimed at enhancing the accuracy and quality of instruction tuning pairs. This pipeline includes improved keyframe extraction, leveraging state-of-the-art large multimodal models (LMMs) for detailed descriptions, and refining the instruction generation strategy. The dataset contains 75K instruction tuning QA pairs, designed to address the limitations of previous datasets.
提供机构:
MBZUAI
原始信息汇总
数据集概述
许可证信息
- 许可证类型:MIT



