1. This repository is used to provide a code base for PyTorch-based training framework. One can easily develop their own model on top of it.
2. Some dirty codes are still there without debugging.
3. UNet-based denoising/segmentation examples are provided based on two open-access datasets
   1. Denoising: AAPM LowDose CT Challenge dataset
   2. Segmentation: StructSeg segmentation dataset.

step by step:

1. customize your own dataset within lib/dataset
2. customize your own specific training workflow within lib/model
3. customize your own architecture within lib/model/module