Brain-IT: Image Reconstruction from fMRI via Brain-Interaction Transformer

Not meaning to sound too critical because it's cool but there's so little information here it's hard to interpret?

I'm kinda wondering about the nature of discrepancies between the inferred and actual images, and what the training process actually entailed.

This is super cool. One step closer to me finally being able to actually record my dreams and show them to other people

wow, imagine (sic!) reading images directly from people minds - that means the inverse is likely achievable - hence ads injected directly in your dreams

Really cool and also a little bit scary.

Incredible - am wondering something similar can be done with EEG