PubMedVision 大规模医学 VQA 数据集

PubMedVision 是一个大规模且高质量的医疗多模态数据集，由深圳市大数据研究院、香港中文大学和 National Health Data Institute 的研究团队于 2024 年创建，包含 130 万个医学 VQA 样本。相关论文成果为「HuatuoGPT-Vision, Towards Injecting Medical Visual Knowledge into Multimodal LLMs at Scale」。

PubMedVision is a large-scale and high-quality medical multimodal dataset developed by research teams from Shenzhen Institute of Big Data, The Chinese University of Hong Kong, and the National Health Data Institute in 2024, which contains 1.3 million medical VQA samples. Its associated research paper is titled "HuatuoGPT-Vision, Towards Injecting Medical Visual Knowledge into Multimodal LLMs at Scale".

搜集汇总

背景与挑战

背景概述

PubMedVision是一个包含130万样本的大规模医学视觉问答数据集，通过PubMed期刊筛选医疗图片和描述，并利用GPT-4V增强图文对齐，构建了10个场景的对话形式。