JunyingWang959/Adv-CPG

Official repository of Adv-CPG (CVPR 2025 oral).

5

stars

106

commits

Python

primary language

Jun 13, 2025

updated

README

Adv-CPG: A Customized Portrait Generation Framework with Facial Adversarial Attacks

How to generate a safe customized portrait?

1School of AI, OPtics and ElectroNics (iOPEN), Northwestern Polytechnical University,
2The University of Hong Kong, 3Institute of Artificial Intelligence (TeleAI), China Telecom
#Corresponding author.

Paper | Github

The overall framework of the proposed Adv-CPG. The framework comprises three key modules: [A] a multi-modal image customizer (MMIC), [B] an ID encryptor (En1), and [C] an encryption enhancer (En2). En1 and En2 enable incremental facial privacy protection, and MMIC achieves fine-grained and personalized portrait generation.

Infer

python infer_adv_cpg.py

Thanks

The code for this work builds upon ConsistentID. We sincerely thank them for their excellent work; please visit their project page for more details. Please maintain the same environment configuration as theirs and directly run infer_adv_cpg.py to execute this work.

Citation

If you found this code helpful, please consider citing:

@InProceedings{Wang_2025_CVPR,
    author    = {Wang, Junying and Zhang, Hongyuan and Yuan, Yuan},
    title     = {Adv-CPG: A Customized Portrait Generation Framework with Facial Adversarial Attacks},
    booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)},
    month     = {June},
    year      = {2025},
    pages     = {21001-21010}
}

Contributors

JackAILab

80 commits

no996no007

9 commits

ssugarwh

6 commits

JunyingWang959/Adv-CPG

Official repository of Adv-CPG (CVPR 2025 oral).

5

stars

106

commits

Python

primary language

Jun 13, 2025

updated

README

Adv-CPG: A Customized Portrait Generation Framework with Facial Adversarial Attacks

How to generate a safe customized portrait?

1School of AI, OPtics and ElectroNics (iOPEN), Northwestern Polytechnical University,
2The University of Hong Kong, 3Institute of Artificial Intelligence (TeleAI), China Telecom
#Corresponding author.

Paper | Github

The overall framework of the proposed Adv-CPG. The framework comprises three key modules: [A] a multi-modal image customizer (MMIC), [B] an ID encryptor (En1), and [C] an encryption enhancer (En2). En1 and En2 enable incremental facial privacy protection, and MMIC achieves fine-grained and personalized portrait generation.

Infer

python infer_adv_cpg.py

Thanks

The code for this work builds upon ConsistentID. We sincerely thank them for their excellent work; please visit their project page for more details. Please maintain the same environment configuration as theirs and directly run infer_adv_cpg.py to execute this work.

Citation

If you found this code helpful, please consider citing:

@InProceedings{Wang_2025_CVPR,
    author    = {Wang, Junying and Zhang, Hongyuan and Yuan, Yuan},
    title     = {Adv-CPG: A Customized Portrait Generation Framework with Facial Adversarial Attacks},
    booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)},
    month     = {June},
    year      = {2025},
    pages     = {21001-21010}
}

Contributors

JackAILab

80 commits

no996no007

9 commits

ssugarwh

6 commits

Languages

Python

91.4%

Cuda

5.2%

C++

2.9%