CelebA HQ Face Identity and Attributes Recognition using PyTorch

This repository provides a CelebA HQ face identity and attribute recognition model using PyTorch.

• This dataset has been first introduced in the official PyTorch implementations for Latent-HSJA.
  • The work is presented at ECCV 2022 Workshop on Adversarial Robustness in the Real World.

Datasets

Facial Identity Recognition Dataset

• There are 307 identities.
  • Each identity has more than 15 images.
• There are 4,263 train images.
• There are 1,215 test images.

Dataset/
    train/
        identity 1/
        identity 2/
        ...
    test/
        identity 1/
        identity 2/
        ...

Face Gender Recognition Dataset

• There are 30,000 gender images.
  • There are 11,057 male images.
  • There are 18,943 female images.
• There are 23,999 train images.
• There are 6,001 test images.

Dataset/
    train/
        male/
        female/
    test/
        male/
        female/

Model Training Examples

Facial Identity Recognition Model

• Training source code
• Testing source code
• Test accuracy: 86.0082%

• Training source code (resolution 256)

Face Gender Recognition Dataset

• Training source code
• Testing source code
• Test accuracy: 98.4003%

• Training source code (resolution 256)

Citation

If this work can be useful for your research, please cite our paper:

@inproceedings{na2022unrestricted,
  title={Unrestricted Black-Box Adversarial Attack Using GAN with Limited Queries},
  author={Na, Dongbin and Ji, Sangwoo and Kim, Jong},
  booktitle={European Conference on Computer Vision},
  pages={467--482},
  year={2022},
  organization={Springer}
}