Harvard-FairSeg
收藏资源简介:
Harvard-FairSeg数据集是由哈佛大学眼科人工智能实验室创建的大规模医学图像分割数据集,专注于公平性学习。该数据集包含10,000个样本,涵盖了年龄、性别、种族、民族、首选语言和婚姻状况等六种敏感属性,旨在通过Segment Anything Model (SAM) 和公平误差边界缩放方法,提高不同身份群体的分割性能公平性。数据集的应用领域主要集中在提高医学图像分割的公平性,特别是在诊断青光眼等眼科疾病时,确保不同人群的准确性和公平性。
The Harvard-FairSeg dataset is a large-scale medical image segmentation dataset developed by the Harvard Ophthalmic AI Lab, with a core focus on fairness-aware learning. It consists of 10,000 samples that incorporate six sensitive attributes: age, gender, race, ethnicity, preferred language, and marital status. The dataset is designed to improve the fairness of segmentation performance across different demographic groups through the Segment Anything Model (SAM) and fair error-bound scaling approaches. Its primary application scenarios center on enhancing the fairness of medical image segmentation, especially during the diagnosis of ophthalmic disorders such as glaucoma, to guarantee consistent accuracy and fairness across diverse patient populations.

- 1FairSeg: A Large-Scale Medical Image Segmentation Dataset for Fairness Learning Using Segment Anything Model with Fair Error-Bound Scaling哈佛大学眼科人工智能实验室 · 2024年



