Keynotes | K&Ts | GACs | Talks | Posters | Search
Poster D in Poster Session D: Wednesday, August 5, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms
re:vision – A community-driven replication initiative for image-related neuroscience
Luca Kämmer1, Josefine Zerbe1,2, Johannes Roth1, Alessandro Thomas Gifford3, Michael Bonner4, Radoslaw Martin Cichy3, Iris Groen5, Margaret M Henderson6, Apurva Ratan Murty7, Martin N Hebart8; 1Max Planck Institute for Human Cognitive and Brain Sciences, 2Justus Liebig University Giessen, 3Freie Universität Berlin, 4Johns Hopkins University, 5University of Amsterdam, 6Carnegie Mellon University, 7Georgia Institute of Technology, 8Justus Liebig Universität Gießen
Presenter: Luca Kämmer
Condition-rich fMRI studies have contributed strongly to our understanding of visual representations in the brain and their alignment with deep neural networks. However, the datasets underlying these studies capture only a part of the visual world, raising questions about the replicability and generalizability of resulting findings. To address this challenge, we introduce re:vision, a community-driven initiative that aims to systematically replicate studies from image-related neuroscience and synthesize the results in a consortium paper. Replications will be conducted using the LAION-fMRI dataset, a new multi-echo 7T fMRI dataset in which five participants completed >30 sessions and each viewed approximately 7,000 images selected to maximally cover visual and semantic space. Beyond replication, the initiative aims to test generalization using out-of-distribution images and semantically defined subsets of the stimulus space. The goal of this large-scale collaborative initiative is to assess which findings are robust beyond their original datasets to establish a stronger empirical foundation for image-related neuroscience and NeuroAI.
Topic Area: Methods, Tools, Theory & Neural Coding