How do I implement “spatial-transpose conv layers” on a time-series of volumes in PyTorch?
I’m trying to implement this 2020 paper: Latent neural source recovery via transcoding of simultaneous EEG-fMRI.
How can I edit this code so that I can insert my own sinogram for image reconstruction?
I have an image that I’m turning into a sinogram to then backproject, but I want to edit this code so that I’m inserting my own sinogram instead (so the true object is unknown).