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
1.49k stars 318 forks source link

Is there a limit to the input image resolution? #12

Closed yinnhao closed 5 years ago

yinnhao commented 5 years ago

github

this error occurred when I feed 1080p video into the network loaded from "EDVR_REDS_deblurcomp_L.pth". everything is ok on lower resolution video. @xinntao

yinnhao commented 5 years ago

'Predeblur' requires the width and height of the image to be a multiple of 8? @xinntao

iPrayerr commented 5 years ago

Same to you. I found that it can only receive frames with resolution 64×64, just like the paper said. I just wanna know if I can change some parameters to accept 1080p video testing.

iPrayerr commented 5 years ago

Don't tell me I can use opencv to resize - it's really a mess

xinntao commented 5 years ago

@hahahaprince , the width and heights of the input image should be a multiple of 4, which used by the predeblur module and the PCD module.

@iPrayerr , it can receive frames whose width and heights are a multiple of 4. If your input video does not meet this requirement, you can pad by zero or reflection. After processing, you can then crop the extra pixels.

yinnhao commented 5 years ago

@hahahaprince , the width and heights of the input image should be a multiple of 4, which used by the predeblur module and the PCD module.

@iPrayerr , it can receive frames whose width and heights are a multiple of 4. If your input video does not meet this requirement, you can pad by zero or reflection. After processing, you can then crop the extra pixels.

okay, thank you!