Skip to content

MindEye2

Reconstruct visual perception from one hour of fMRI data

Visit official site

What MindEye2 is

MindEye2 is described as a project showing how visual perception can be reconstructed with one hour of fMRI data, after a model is pretrained on several subjects and fine-tuned on a new one.

What you can do with it

  • Map fMRI data to a visual representation
  • Pretrain a model across several subjects
  • Fine-tune the model on a new subject
  • Reconstruct visual perception from the data

Model notes

  • The approach uses a shared subject model
  • A pretraining stage comes before fine-tuning
  • One hour of data is described for a new subject

Pros & cons

✓ What we like

  • Only one hour of fMRI data is described for a new subject
  • Pretraining covers several subjects before fine-tuning

! What to watch out for

  • The method requires a fine-tuning step for a new subject
  • The reconstruction is tied to fMRI data mapping

FAQ

How much data is used?

The material describes using one hour of fMRI data for a new subject.

What is reconstructed?

Visual perception is reconstructed from fMRI data through the shared subject model.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review