Voxels Activity Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://sites.google.com/view/lavca-llm/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了用于评估LaVCa方法在生成描述像素选择性的自然语言标题方面性能的体素级大脑活动数据。此外，该数据集还用于将LaVCa方法与现有的BrainSCUBA等方法进行比较，重点关注像素间和像素内的多样性度量。该数据集基于训练数据上准确度最高的前5000个体素进行筛选，其任务是对体素的大脑活动预测和标题生成进行评估。
This dataset contains voxel-level brain activity data used to evaluate the performance of the LaVCa method in generating natural language captions that describe pixel-wise selectivity. Furthermore, this dataset is also employed to compare the LaVCa method with existing approaches such as BrainSCUBA, with a focus on inter-pixel and intra-pixel diversity metrics. This dataset is screened based on the top 5000 voxels with the highest accuracy on the training data, and its evaluation tasks aim to assess brain activity prediction and caption generation for voxels.
提供机构：
Research team associated with LaVCa
