PathMMU
收藏资源简介:
PathMMU是由浙江大学等机构创建的大型多模态病理学基准数据集,包含33,428个多模态多项选择题和24,067张来自不同来源的图像。数据集通过GPT-4V的先进能力,利用超过30,000对图像标题来丰富描述并生成相应的问题和答案。PathMMU旨在通过严格的专家验证,为大型多模态模型提供一个高质量的评估平台,以推动病理学领域的AI发展。
PathMMU is a large-scale multimodal pathology benchmark dataset developed by Zhejiang University and other institutions. It contains 33,428 multimodal multiple-choice questions and 24,067 images sourced from various origins. The dataset enriches image descriptions and generates corresponding questions and answers by leveraging the advanced capabilities of GPT-4V, with over 30,000 image-caption pairs employed. PathMMU aims to provide a high-quality evaluation platform for large multimodal models through rigorous expert validation, so as to promote the advancement of AI in the field of pathology.



