BOLD5000

BOLD5000是由卡内基梅隆大学创建的一个大型公共fMRI数据集，包含近5000张真实世界场景的图像。该数据集不仅在规模上远超以往的fMRI研究，还通过包含来自标准计算机视觉数据集（如SUN、COCO和ImageNet）的图像，确保了数据多样性和重叠性。BOLD5000的设计采用慢事件相关fMRI方法，使得能够精细探索广泛视觉特征、类别和语义的神经表征。该数据集旨在促进人类和计算机视觉领域之间的更大合作，实现Marr的单一视觉科学梦想。

BOLD5000 is a large-scale public fMRI dataset developed by Carnegie Mellon University, containing nearly 5,000 images of real-world scenes. It not only surpasses previous fMRI studies in scale, but also ensures data diversity and overlap by incorporating images from standard computer vision datasets such as SUN, COCO, and ImageNet. Adopting the slow event-related fMRI paradigm, BOLD5000 enables fine-grained exploration of neural representations across a wide range of visual features, categories and semantics. This dataset aims to foster greater collaboration between the fields of human and computer vision, realizing Marr's vision of a unified science of vision.