Algonauts 2021 (How the Human Brain Makes Sense of a World in Motion)

Algonauts 数据集提供人类大脑对一组 1,102 个 3 秒长的日常事件视频片段的反应。用功能磁共振成像 (fMRI) 测量大脑反应。 fMRI 是一种广泛使用的脑成像技术，具有高空间分辨率，可测量与神经反应相关的血流变化。拆分：训练：训练集由 1,000 个视频剪辑和相关的大脑反应组成。大脑反应在这里以两条轨道提供，对应于 Algonauts 挑战中的两条独立轨道。在第一首曲目中，提供的大脑反应来自一组已知在视觉感知中起关键作用的特定感兴趣区域 (ROI)。这些 ROI 从早期和中期视觉皮层（V1、V2、V3 和 V4）开始，并延伸到优先响应所有对象或特定类别（身体-EBA；面部-FFA、STS；对象- LOC；场景 - PPA）。在第二个曲目中，提供的大脑反应来自整个大脑中选定的体素，显示出对视频的可靠反应。测试：测试集由102个短视频组成。相关的大脑反应将在以后发布。有关更多详细信息，请参见此处。

The Algonauts Dataset provides human brain responses to a set of 1,102 3-second-long everyday event video clips. Brain responses are measured using functional magnetic resonance imaging (fMRI), a widely used brain imaging technique with high spatial resolution that measures blood flow changes associated with neural activity. Split details: The training set consists of 1,000 video clips and their corresponding brain responses. Here, brain responses are provided in two tracks, corresponding to two separate tracks in the Algonauts Challenge. In the first track, the provided brain responses are derived from a set of specific regions of interest (ROIs) known to play critical roles in visual perception. These ROIs span early and mid-level visual cortices (V1, V2, V3, and V4) and extend to areas preferentially responding to all objects or specific categories: body-selective EBA; face-selective FFA and STS; object-selective LOC; and scene-selective PPA. In the second track, the provided brain responses are derived from selected voxels across the entire brain that exhibit reliable responses to videos. The test set consists of 102 short videos. Its corresponding brain responses will be released at a later date. For more details, please see here.

搜集汇总

背景与挑战

背景概述

Algonauts 2021数据集旨在研究人类大脑如何理解动态世界，提供了1,102个日常事件视频片段及其功能磁共振成像（fMRI）测量的大脑反应。训练集包含1,000个样本，涵盖特定视觉区域和全脑可靠体素的两条轨道，而测试集由102个视频组成，其大脑反应暂未发布。该数据集由多个机构于2021年联合发布，用于推动神经科学和计算机视觉的交叉研究。