Hi, I’ve tried all of the dnn_superres models, and although they work well with the example images such as the peppers etc, with my own images they just appear completely blurry, and are virtually identical or more blurry than bicubic.
Could anyone offer any tips why this could be? Perhaps I need to modify the input images in some way for better results?
CUBIC 4x (actually more detail retained!)
RealESRGAN Sharp but, not correct.