Awesome

Robust-Wide: Robust Watermarking Against Instruction-Driven Image Editing (ECCV2024)

Official implementation of Robust-Wide: Robust Watermarking Against Instruction-Driven Image Editing.

Train

1. Download the data and put them into the data dir ./data.

2. Configure the train script and then run it.

bash train.sh

Inference

1. Put your original image in ./examples.

2. Download the checkpoints and put them in ./checkpoints.

3. Configure the inference script and then run it.

bash inference.sh

Acknowledgements

This code builds on the code from the diffusers library.

Cite

If you find this repository helpful, please cite the following:

@inproceedings{hu2025robust,
  title={Robust-wide: Robust watermarking against instruction-driven image editing},
  author={Hu, Runyi and Zhang, Jie and Xu, Ting and Li, Jiwei and Zhang, Tianwei},
  booktitle={European Conference on Computer Vision},
  pages={20--37},
  year={2025},
  organization={Springer}
}