Result for training SR-ResNet network to perform single-image super resolution.
The SR-ResNet mocel is trained on DIV2K dataset and inference is run on BSD test set.
To compare only selected loss functions, tick their names and click submit below. Click on any restored image to see its original version.