DFCAN/DFGAN could get an amazing result in SIM super resolution reconstruction in optical microscopy.
Code is basically a pytorch implementation of DFCAN/DFGAN(https://www.nature.com/articles/s41592-020-01048-5),and in reference to the tensorflow/keras implementation from https://github.com/qc17-THU/DL-SR.
All credit goes to the authors of DFCAN/DFGAN, listed in the paper“Evaluation and development of deep neural networks for image super-resolution in optical microscopy”.
This implementation is without the process of prctile_norm and recovery.And as a result,the tail in the coding has no sigmoid function. The expected high resolution picture will be acquired from the model directly.
1.pytorch 1.7.1 or higher