- This is a reimplementation of paper Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising using Matlab Built-in deeplearning toolbox
- The original version uses Matconvnet
- Matlab R2018a only
- set up your configurations as in main*.m
- use
Project.run_project(opts)to run the project
- About time: Due to the implementation of Matlab BatchNorm layers, which doesn't support testing before training is finished, I have to make an trade-off (i.e., force finalizing training for each epoch, which is time-consuming) in order to test the performance(PSNR and SSIM) of each snapshot of the network, hence the total training time is about 10 hours for 50 epochs, which is longer than the original ones(6 hours)
- Fine-tuned network models
- more experiments on different training options
Have fun fine-tuning :)