- multiprocessing for pre-processing NIFTI volumes
- metrics calculation and logging
- adding noise to simulate noisy images for denoising
using whole-brain inference
- Generate LR whole brain volume
- Infer them to get whole brain SR volume
- Suggested to be used on
device=cpu
when your GPU mem is low
or using individual-patch inference
- Take a whole brain volume, downsample to LR
- Patch them into overlapped volumes
- Infer individual LR patches to get SR patches
- Assemble them to get a SR whole brain volume