Using GANs for denoising images: https://openaccess.thecvf.com/content/ACCV2020/papers/Tran_GAN-based_Noise_Model_for_Denoising_Real_Images_ACCV_2020_paper.pdf
- Add varying types of noise after each epoch
- Show training and validation loss over epochs
- Get close to -20db
- Above 19.2
- K-fold cross validation
- Bioimaging techniques (Homework 5)
Denoising architecture: https://arxiv.org/pdf/2208.14337.pdf https://stanford.edu/class/ee367/Winter2016/Chaudhari_Report.pdf
Paper on different denoising architecture used for MRI images: https://www.sciencedirect.com/science/article/pii/S0730725X19304643
How to use Optuna (example): https://optuna.org
Pytorch:
- For training set you should shuffle, not for the validation set
- One loader for each dataset
- Does the backpropogation internally
- self.conv2D = nn.Conv2D() # double check syntax