Methods

Important

These pages summarize the Methods section of the AVS dataset paper: Sulewski, Amme, König, Hebart & Kietzmann (2026), “Active Visual Semantics: A large-scale MEG and eye-tracking dataset for understanding visual intelligence in action”, currently available as a preprint (arXiv:2609.01055, see Citation). Figures cited here are taken directly from that paper. If a peer-reviewed version is published later, this page will be updated to match – treat this as a working summary, not a substitute for the published paper.

The AVS dataset was collected to study active vision: brain activity during self-directed scene exploration, rather than passive viewing with enforced central fixation. These pages cover, in order:

  1. Participants – who was recorded, and the session schedule

  2. Stimuli – the natural scene stimulus set and how it was selected

  3. MEG Acquisition – the MEG system and head-stabilization setup

  4. Eye Tracking – the eye-tracking system, calibration, and event detection

  5. MEG Preprocessing – filtering, ICA artifact removal, and epoching

  6. Fixation Object Labeling – mapping fixations to MS-COCO / COCO-Stuff object categories

  7. Source Reconstruction – forward modeling and beamforming

  8. Semantic Captioning Task – the verbal scene-description task

For how to run the corresponding processing steps with pyAVS, see AVSComposer Guide and Tutorials.