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
207 stars 28 forks source link

Training config file for HuggingFace Demo #18

Open yx-chan131 opened 1 year ago

yx-chan131 commented 1 year ago

Hi, great work by the way! May I know the training configurations for the HuggingFace Demo?

XinYu-Andy commented 1 year ago

Sorry for the late reply - my research has shifted away from this area, so I haven't been keeping up with this repo much.

About the demo config: it's pretty straightforward. You'll just need to write a new dataloader that can handle merging and loading multiple datasets. For handling large images, go for cropping, and for the TIP2018 dataset, resize it firstly and then do cropping. All the specific details for these implementations are mentioned in the paper, so you can check them out there. No extra parameter tweaking needed beyond that. Hope this helps!