NAFNet
NAFNet copied to clipboard
megvii-research
Performance of deblurring images which has high gaussian and median blur spots
Hi team,
Great and thanks for the work team! My dataset involves a mix up of gaussian blur, fog, median blur and few cases of motion blur, Infact in some cases, the single input image will have mix up of the above-mentioned blur. I trained Nafnet for 200,000 iterations with pair-to-pair input images, which is around 3k. But could see very less deblurring happening that too only in few cases. I also tired giving input images without sub cropping it, but couldn't find improvements. Is there something which I'm missing to get better results. Kindly guide with this! Thanks!
Hello, I am preparing to train this network to perform Gaussian blur deblurring. Can you please take a look at your code for training this network to deblur? I can't find a clue, thank you very much. In addition, you mentioned that this network does not seem to be very effective at removing Gaussian blur. Can you take a look at your renderings?
