Related Experiment Video
Updated: Jul 29, 2025
06:02
Topographical Estimation of Visual Population Receptive Fields by fMRI
Published on: February 3, 2015
9.3K
Reconstructing seen images from human brain activity via guided stochastic search
Reese Kneeland1, Jordyn Ojeda1, Ghislain St-Yves2
1Department of Computer Science, University of Minnesota, Minneapolis, MN 55455 USA.
Arxiv
|May 19, 2023
Summary
This study introduces a new method using AI diffusion models to reconstruct images from brain activity (fMRI). This approach improves visual reconstruction accuracy and offers insights into brain representation diversity.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Computer Vision
Background:
- Visual reconstruction algorithms interpret brain activity by mapping it to pixels.
- Previous methods relied on brute-force searches through image libraries and encoding models.
Conclusions:
- Conditional generative diffusion models offer a powerful and efficient approach to visual brain activity reconstruction.
- The observed systematic differences in convergence time provide a new metric for understanding visual cortex organization and representational diversity.
