zhuyr97 / HGRR

HGRR

This is the official implementation of the TNNLS 2023 paper "Hue Guidance Network for Single Image Reflection Removal".

Implementation

1. Prepare the training data as follows:

     |--training_data
          |--JPEGImages  # VOC2012 images for synthesis the reflection data 
          |--real_train  # real world reflection images and corresponding GT from previous methods, eg, real, nature 
              |--blended   #   reflection images
              |--transmission_layer   #  ground truth (GT)
          
   

2. training phase

   python train_HGRR.py --name train_DB_wboosting --hyper --inet HGRR_wboosting1128 --lambda_vgg 0.02 --lambda_newloss 50 --lambda_newloss_H 20

3. evaluation phase

   python test_HGRR.py --name test -r --inet HGRR_wboosting1128 --icnn_path best.pt --hyper

Results & Pre-trained model

Results of Reflection Benchmarks

Pre-trained model

Concat

If you have any problem, please feel free to contact me ([e-mail]).

Acknowledgments

• Our codes are inspired by ERRNet