A Cognitive Process-Inspired Architecture for Subject-Agnostic Brain Visual Decoding

Jingyu Lu, Haonan Wang, Qixiang Zhang, Xiaomeng Li · 2025

Read A Cognitive Process-Inspired Architecture for Subject-Agnostic Brain Visual Decoding by Jingyu Lu, Haonan Wang, Qixiang Zhang, Xiaomeng Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-02565.txt.

Description

Jingyu Lu, Haonan Wang, Qixiang Zhang, and Xiaomeng Li reconstruct what a person is seeing directly from brain scans, without first training on that person's own data. Decoding vision from functional MRI usually demands more than twelve hours of recordings from each subject, which blocks clinical use. Their architecture, VCFlow, copies the organisation of the human visual system, separating early visual cortex features from the ventral stream, which recognises objects, and the dorsal stream, which handles space and motion.

A contrastive learning strategy at the feature level pulls out semantic representations that stay stable across different people's brains, the key to working on unseen subjects. The trade is modest: about seven percent average accuracy given up, in exchange for generating each reconstructed video in ten seconds with no per-subject retraining at all. The authors pitch the system as a fast, clinically scalable route to brain-based visual decoding.

Catalog record

CREATORJingyu Lu, Haonan Wang, Qixiang Zhang, Xiaomeng Li
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-02565

Browse the full library · About the Tower of Babel