Showing 1–1 of 1 results for author: Zangos, C

Efficient Multi Subject Visual Reconstruction from fMRI Using Aligned Representations

Authors: Christos Zangos, Danish Ebadulla, Thomas Christopher Sprague, Ambuj Singh

Abstract: This work introduces a novel approach to fMRI-based visual image reconstruction using a subject-agnostic common representation space. We show that the brain signals of the subjects can be aligned in this common space during training to form a semantically aligned common brain. This is leveraged to demonstrate that aligning subject-specific lightweight modules to a reference subject is significantl… ▽ More This work introduces a novel approach to fMRI-based visual image reconstruction using a subject-agnostic common representation space. We show that the brain signals of the subjects can be aligned in this common space during training to form a semantically aligned common brain. This is leveraged to demonstrate that aligning subject-specific lightweight modules to a reference subject is significantly more efficient than traditional end-to-end training methods. Our approach excels in low-data scenarios. We evaluate our methods on different datasets, demonstrating that the common space is subject and dataset-agnostic. △ Less

Submitted 2 May, 2025; originally announced May 2025.