Awesome
ResidualDenseNetWork-CycleGAN-SuperResolution
Using Residual Dense Network as one of Generator and define another RDN within a downsampler
Pytorch implement:
Residual Dense Network for Image Super-Resolution
Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
Different with the paper 1, I use there RDBs(Residual dense block), every RDB has three dense layers. So ,this is a sample implement the RDN(Residual Dense Network) proposed by the author.
Different with the paper 2, I use paired training data to constrain the convert performance.
Requirements
- python3.6
- pytorch >= 1.0
- torchvision >= 2.1
- opencv
Dateset
you need prepare DIV2K dataset Download the High Resolution and (2x/3x/4x) Low Resolution put them in ./DIV2K