Closed StevenLDZ closed 5 years ago
thanks for your response,i’ve tested that with my own aligned blur dataset with 128x128 ,point dataroot、valDataroot to the dataset path,but the results is not as good as the paper,maybe the input is not correct?
Is the input is just blurry image of size 128x128? Did you re-train the network or using the pre-trained model?
yes,aligned 128x128 faces after face detection,i used the pre-trained model
you can send me link to sample images of your test dataset to ryasarl1@jhu.edu. I will have a look at it and get back you
hi ,have you solved the problem of bad results?
I didn't receive any email from StevenLDZ. But the network is fine it's giving me the good results. Let me know if you face the same issue of bad results
you can follow the instructions in the section "To test UMRL" to test the code datasets provided by authors of Ziyi et al. If you want to run test on your own data. Just provided path to dataset with images naming like 000001.png run the following command python test_face_deblur.py --dataroot ./facades/github/ --valDataroot --netG ./pretrained_models/Deblur_epoch_Best.pth