search

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

HI, would you please help me with the question bellow? How to train the deblur and deblurcomp modles. Do they the same with the SR setting? Waiting for your help online. Thanks! #87

Open zcpchaos opened 4 years ago

xinntao commented 4 years ago

Almost the same but with pretraining the Predeblur module first.

zcpchaos commented 4 years ago

Thanks again! But how to train the Predeblur module? Would you please provide the scripts and yml settings?

xinntao commented 4 years ago

We first train a small image deblur network (the same network architecture as the PreDeblur part in EDVR, followed by a final image output; we use BasicSR codes to pretrain the image deblur) and then initialize the PreDeblur part in EDVR.

We will sort out a clean and tidy training procedure later.

wangyiwei0306 commented 4 years ago

@xinntao Hi, my debluring performance is also bad. Have you finished sorting out the clean and tidy training procedure. Thank you very much for your help

RenMing-Huang commented 2 years ago

Thanks! I solved it finally