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

Source code or reproduction code of DMCNN,MBCNN and MDDM models #14

Closed Thothene closed 1 year ago

Thothene commented 1 year ago

Hello！ I recently wanted to retrain other demoir´eing models, but found that DMCNN did not give the source code but only the code that others reproduced, while MBCNN and MDDM only had the test code. Would you like to ask if you have their source code or just retrained the model based on their test code? Thank you! DMCNN is from Moir´e photo restoration using multiresolution convolutional neural networks. MBCNN is from Image demoireing with learnable bandpass filters. MDDM is from Multi-scale Dynamic Feature Encoding Network for Image Demoireing.

XinYu-Andy commented 1 year ago

Hello！ I recently wanted to retrain other demoir´eing models, but found that DMCNN did not give the source code but only the code that others reproduced, while MBCNN and MDDM only had the test code. Would you like to ask if you have their source code or just retrained the model based on their test code? Thank you! DMCNN is from Moir´e photo restoration using multiresolution convolutional neural networks. MBCNN is from Image demoireing with learnable bandpass filters. MDDM is from Multi-scale Dynamic Feature Encoding Network for Image Demoireing.

For MBCNN, we followed its official TensorFlow code and reproduced it using PyTorch. For DMCNN, we modified (solve some bug) the unofficial code and retrained it. For MDDM, we used the official code and retrained it.

Thothene commented 1 year ago

Thanks a lot，I will try later based on the help you provide.