SemiHVision
收藏资源简介:
SemiHVision数据集由马萨诸塞大学阿默斯特分校信息与计算机科学学院创建,旨在通过结合人类注释和自动化增强技术,提升医学多模态模型的性能。该数据集包含1400万条样本,涵盖多种医学图像和文本描述,主要来源于PubMed数据集。数据集的创建过程包括对原始数据进行筛选和分类,使用GPT-4o进行文本再生和标准化处理。SemiHVision数据集主要应用于医学领域的多模态模型训练和评估,旨在提高模型在实际临床任务中的诊断推理能力。
The SemiHVision dataset was developed by the School of Information and Computer Sciences at the University of Massachusetts Amherst, with the goal of improving the performance of medical multimodal models by integrating human annotations and automated augmentation techniques. Comprising 14 million samples, this dataset covers diverse medical images and textual descriptions, and is primarily derived from the PubMed dataset. The dataset development workflow involves screening and categorizing the raw data, followed by text regeneration and standardization processing using GPT-4o. The SemiHVision dataset is primarily utilized for training and evaluating multimodal models in the medical domain, with the objective of enhancing the models' diagnostic reasoning capabilities in real-world clinical tasks.
SemiHVision
数据集概述
- 名称: SemiHVision
数据集详情
- 描述: 该数据集的README文件内容为空,无法提供更多详细信息。

- 1SemiHVision: Enhancing Medical Multimodal Models with a Semi-Human Annotated Dataset and Fine-Tuned Instruction Generation马萨诸塞大学阿默斯特分校信息与计算机科学学院 · 2024年



