search

CVMI-Lab / UHDM

(ECCV2022) This is the official PyTorch implementation of ECCV2022 paper: Towards Efficient and Scale-Robust Ultra-High-Definition Image Demoireing
Apache License 2.0
191 stars 25 forks source link

How does it compare to this project(VideoDemoireing), #7

Closed zhanghongyong123456 closed 1 year ago

zhanghongyong123456 commented 1 year ago

  1. i test my pic by https://github.com/CVMI-Lab/VideoDemoireing I test my pic ,The results are very poor. image image

  2. this pro compare with VideoDemoireing ,what is the difference? 1. quality? 2. One（VideoDemoireing） for video and one（this project） for pictures?， i see you promat： image

XinYu-Andy commented 1 year ago
  1. i test my pic by https://github.com/CVMI-Lab/VideoDemoireing I test my pic ,The results are very poor. image image
  2. this pro compare with VideoDemoireing ,what is the difference? 1. quality? 2. One（VideoDemoireing） for video and one（this project） for pictures?， i see you promat： image

Hi,

The network of video demoireing is different with this repo. You could try this repo for image demoireing. The key in video demoireing is the relation-based loss to solve the temporary consistency.

zhanghongyong123456 commented 1 year ago
  1. i test my pic by https://github.com/CVMI-Lab/VideoDemoireing I test my pic ,The results are very poor. image image
  2. this pro compare with VideoDemoireing ,what is the difference? 1. quality? 2. One（VideoDemoireing） for video and one（this project） for pictures?， i see you promat： image

Hi,

The network of video demoireing is different with this repo. You could try this repo for image demoireing. The key in video demoireing is the relation-based loss to solve the temporary consistency.

I find this project have 60fps 4K on 3090, That's consistent with the video stream removal morire pattern,but (VideoDemoireing pro )The effect is worse than this project. What should I do, for VideoDemoireing pro 4K 60fps? Looks like a small resolution

XinYu-Andy commented 1 year ago
  1. i test my pic by https://github.com/CVMI-Lab/VideoDemoireing I test my pic ,The results are very poor. image image
  2. this pro compare with VideoDemoireing ,what is the difference? 1. quality? 2. One（VideoDemoireing） for video and one（this project） for pictures?， i see you promat： image

Hi, The network of video demoireing is different with this repo. You could try this repo for image demoireing. The key in video demoireing is the relation-based loss to solve the temporary consistency.

I find this project have 60fps 4K on 3090, That's consistent with the video stream removal morire pattern,but (VideoDemoireing pro )The effect is worse than this project. What should I do, for VideoDemoireing pro 4K 60fps? Looks like a small resolution

I think this is a future work, i.e., video demoireing on 4K resolution. The key problem is there not exists a 4K moire-video dataset for training currently, or we can combine our image demoireing framework and relation-based loss to train the desired model.