jiangsutx / SRN-Deblur

Repository for Scale-recurrent Network for Deep Image Deblurring
http://www.xtao.website/projects/srndeblur/srndeblur_cvpr18.pdf
MIT License
709 stars 184 forks source link

About the number of training iterations #59

Closed WongJianyang closed 3 years ago

WongJianyang commented 3 years ago

The default training is 4000 epochs, which is too time-consuming. If I change to 500 epochs, will the test result drop too much?

jiangsutx commented 3 years ago

Yes, the performance will drop. But I think 500 epochs will be enough to see a difference between output and input. I mean sharper and clearer images.

You may finetune using existing models.