Open Lcjgh opened 1 year ago
Sorry but that is really not the case. Best is if you start with the paper and ask questions accordingly.
In brief - G learns the image-specific kernel of the input image, then we inject the image+kernel to the ZSSR work (or any SR model) that produces HR image.
Thank you very much for your answer.
Thank you for sharing the code. Could you tell me what is the whole operation process? The real LR images are used to train the G network, and then the trained G network is used to process the real HR images to generate the synthetically LR images, and finally the HR-LR image pairs are formed and sent to the SR network.