Locally Adaptive Adversarial Color Attack

This project is for the LAACA we proposed in the paper "Artwork Protection Against Neural Style Transfer Using Locally Adaptive Adversarial Color Attack", which is accepted by 50th European Conference on Artificial Intelligence (ECAI 2024).

The dataset folder contains the content images and style images.

The main implementation is in laaca.py.

To run the method with some content images and style images, please use the following bash script:

python main.py

the details for parameters can be found via:

python main.py --help

Our proposed Image Quality Assessment (IQA), Aesthetic Color Distance Metric (ACDM), can be found in this repo.

Environment

numpy==2.0.0
pandas==1.5.2
Pillow==9.2.0
torch==2.0.0
torchvision==0.15.1
tqdm==4.65.0

Cite

@incollection{guo2024artwork,
  title={Artwork protection against neural style transfer using locally adaptive adversarial color attack},
  author={Guo, Zhongliang and Dong, Junhao and Qian, Yifei and Wang, Kaixuan and Li, Weiye and Guo, Ziheng and Wang, Yuheng and Li, Yanli and Arandjelovi{\'c}, Ognjen and Fang, Lei},
  booktitle={ECAI 2024},
  pages={1414--1421},
  year={2024},
  publisher={IOS Press}
}