DENet: Disentangled Embedding Network for Visible Watermark Removal

This is official implementation for paper DENet: Disentangled Embedding Network for Visible Watermark Removal [AAAI2023 Oral]

Dataset preparation

|--data
|--|--LOGO
   |--|--10kmid
   |--|--10kgray
   |--|--10khigh

Pretrained Model

	PSNR	SSIM	LPIPS
LOGO-L	44.24	0.9954	0.54
LOGO-H	40.83	0.9919	0.89
LOGO-Gray	42.60	0.9944	0.53

Installation

pip install -r requirements.txt

Training

Train on LOGO-H

bash scripts/train_contrast_attention_on_logo_high.sh 

Train on LOGO-L

bash scripts/train_contrast_attention_on_logo_mid.sh 

Train on LOGO-Gray

bash scripts/train_contrast_attention_on_logo_gray.sh

Testing

Test on LOGO-H

bash scripts/test_LOGO_10khigh.sh

Test on LOGO-L

bash scripts/test_LOGO_10kmid.sh

Test on LOGO-Gray

bash scripts/test_LOGO_10kgray.sh

Acknowledgement

This code is mainly based on the previous work SLBR.