About the parameter ‘niter’ in train_EDVR_m.yml

xinntao / EDVR

Winning Solution in NTIRE19 Challenges on Video Restoration and Enhancement (CVPR19 Workshops) - Video Restoration with Enhanced Deformable Convolutional Networks. EDVR has been merged into BasicSR and this repo is a mirror of BasicSR.

https://github.com/xinntao/BasicSR

1.48k stars 320 forks source link

About the parameter ‘niter’ in train_EDVR_m.yml #139

Open wenchen4321 opened 4 years ago

wenchen4321 commented 4 years ago

Thank you for your codes! I want to ask you why set the parameter ‘niter’ in train_EDVR_m.yml for 600000. I know it is the total number of iteration of a epoch. But I found that in ESRGAN is 400000. Why they are different ?

TomArrow commented 4 years ago

Unless I'm mistaken, niter isn't the number of iterations per epoch, but rather the total number of iterations for the entire training. The iterations per epoch are determined by batch size and the size of your dataset iirc.