Awesome

Visible Watermark Removal via Self-calibrated Localization and Background Refinement

Introduction

This is the official code of the following paper:

Visible Watermark Removal via Self-calibrated Localization and Background Refinement[1] <br>Jing Liang¹, Li Niu¹, Fengjun Guo², Teng Long² and Liqing Zhang¹ <br>¹MoE Key Lab of Artificial Intelligence, Shanghai Jiao Tong University <br>²INTSIG<br> (ACM MM 2021 | Bibtex)

SLBR Network

Here is our proposed SLBR(Self-calibrated Localization and Background Refinement). Top row depicts the whole framework of SLBR and bottom row elaborates the details of our proposed three modules.

Quick Start

Install

• Install PyTorch>=1.0 following the official instructions
• git clone https://github.com/bcmi/SLBR-Visible-Watermark-Removal.git
• Install dependencies: pip install -r requirements.txt

Data Preparation

In this paper, we conduct all of the experiments on the latest released dataset CLWD[2] and LVW[3]. You can contact the authors of LVW to obtain the dataset.

Train and Test

• How to train and test my model?

  We provide an example of training and a test bash respectively:scripts/train.sh, scripts/test.sh

  Please specify the checkpoint save path in --checkpoint and dataset path in--dataset_dir.

• How to test on my data?

  We also provide an example of a custom data test bash: scripts/test_custom.sh And you can further tailor test_custom.py to meet your demands. For the best performance, it is better to finetune on your dataset since our training data size is set as 256x256.

Pretrained Model

Here is the model trained on CLWD dataset:

• Google Drive

• OneDrive

Visualization Results

We also show some qualitative comparision with state-of-art methods:

Acknowledgements

Part of the code is based upon the previous work SplitNet[4].

Citation

If you find this work or code is helpful in your research, please cite:

@inproceedings{liang2021visible,
  title={Visible Watermark Removal via Self-calibrated Localization and Background Refinement},
  author={Liang, Jing and Niu, Li and Guo, Fengjun and Long, Teng and Zhang, Liqing},
  booktitle={Proceedings of the 29th ACM International Conference on Multimedia},
  pages={4426--4434},
  year={2021}
}

Resources

We have summarized the existing papers, codes, and datasets on visible watermark removal in the following repository: https://github.com/bcmi/Awesome-Visible-Watermark-Removal

Reference

[1] Jing Liang, Li Niu, Fengjun Guo, Teng Long and Liqing Zhang. 2021. Visible Watermark Removal via Self-calibrated Localization and Background Refinement. In Proceedings of the 29th ACM International Conference on Multimedia. download

[2] Liu, Yang and Zhu, Zhen and Bai, Xiang. 2021. WDNet: Watermark-Decomposition Network for Visible Watermark Removal. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision.

[3] Danni Cheng, Xiang Li, Wei-Hong Li, Chan Lu, Fake Li, Hua Zhao, and WeiShi Zheng. 2018. Large-scale visible watermark detection and removal with deep convolutional networks. In Chinese Conference on Pattern Recognition and Computer Vision. 27–40.

[4] Xiaodong Cun and Chi-Man Pun. 2020. Split then Refine: Stacked Attentionguided ResUNets for Blind Single Image Visible Watermark Removal. arXiv preprint arXiv:2012.07007 (2020).