rmpku / CIN

Paper 'Towards Blind Watermarking: Combining Invertible and Non-invertible Mechanisms' in ACM Multimedia '22.
49 stars 3 forks source link

Paper

Available on arXiv:
https://arxiv.org/abs/2212.12678
or ACM DL (The supplements are at the end of the PDF file, which contains the description of the CIN, training details, and noise setting):
https://dl.acm.org/doi/abs/10.1145/3503161.3547950

Introduction

Framework Visualization invertibleNet

Dataset Preparation

COCO2017:
Lin, Tsung-Yi, et al. "Microsoft coco: Common objects in context." European conference on computer vision. Springer, Cham, 2014.

DIV2K:
Agustsson, Eirikur, and Radu Timofte. "Ntire 2017 challenge on single image super-resolution: Dataset and study." Proceedings of the IEEE conference on computer vision and pattern recognition workshops. 2017.

Environment

nvidia=3080
cuda=11.1
python=3.8.3
torch=1.13.0
torchvision=0.14.0
opencv-python=4.6.0.66
kornia=0.6.8
colormath=3.0.0
pyyaml=6.0
importlib-metadata=5.1.0

Pretrained model - "combined noises"

Training with Noise pool:
{'Identity', 'JpegTest', 'Crop', 'Cropout', 'Resize', 'GaussianBlur', 'Salt*Pepper', 'GaussianNoise', 'Dropout', 'Brightness', Contrast', 'Saturation', 'Hue'}
When testing:
you only need to modify the noise-option in /codes/options/opt.yml/noise/option
Google Cloud link:
https://drive.google.com/file/d/1wqnqhPv92mHwkEI4nMh-sI5aDgh-usr7/view?usp=share_link

Citation

If you find this work useful, please cite our paper:

ACKNOWLEDGMENTS

This work was supported by National Key R&D Program of China 2021ZD0109802 and National Science Foundation of China 61971047.

Contact

If you have any questions, please contact rui_m@stu.pku.edu.cn or post them in the https://github.com/rmpku/CIN/issues.