Incorporate additional open source datasets into the pretraining step - Githubissues

MedARC-AI / MindEye_Imagery

MIT License

4 stars 0 forks source link

Incorporate additional open source datasets into the pretraining step #4

Open reesekneeland opened 5 months ago

reesekneeland commented 5 months ago

This issue will likely follow from https://github.com/MedARC-AI/MindEye_Imagery/issues/8

The goals for incorporating more data are:

Expanding the prior to encompass simple stimuli
Pretrain on imagery trials from other datasets
Pretrain on more vision data to see general improvements in the models.

Some candidate datasets are below, these will probably have varying levels of utility for this project. Generic Object Decoding Deep Image Reconstruction NeuroMod CMI Neuroimaging Dataset