This software is part of MAI-BIAS; a low-code toolkit for fairness analysis and mitigation, with an accompanying suite of coding tools. Our ecosystem operates in multidimensional and multi-attribute settings (safeguarding multiple races, genders, etc), and across multiple data modalities (like tabular data, images, text, graphs). Learn more here.

FLAC is a state-of-the-art method for learning fair visual representations by reducing the influence of hidden biases — without requiring protected attribute labels.

It enables teams to build AI systems that are:

• More robust
• More trustworthy
• Better aligned with fairness and regulatory requirements

Unlike traditional approaches, FLAC does not rely on demographic annotations (e.g., gender, race), making it highly applicable in real-world scenarios where such labels are unavailable or restricted.

• AI researchers working on fairness and representation learning
• ML engineers building production-ready vision systems
• Decision-makers & product owners aiming to deploy fair and compliant AI

• ✅ No sensitive labels required → avoids privacy and legal risks
• ✅ Fairness by design → reduces bias during training
• ✅ Maintains strong performance while improving fairness
• ✅ Works in real-world conditions with incomplete data

• Minimizes dependence between learned features and hidden biases
• Does not require protected attribute annotations
• Uses a sampling strategy to highlight underrepresented data
• Leverages a bias-capturing classifier to guide training
• Theoretically grounded with guarantees for fair representations

python = 3.6  
pytorch = 1.10.1  

We also provide the environment file used in our experiments:

conda env create -f pytorch36_environment.yml

Download and extract:

• UTKFace
• CelebA
• ImageNet
• ImageNet-A

Place them in a data directory.

Download from: https://github.com/gsarridis/FLAC/releases/tag/bcc

Place them under: bias_capturing_classifiers/

python train_biased_mnist.py --corr 0.99 --alpha 110  
python train_biased_mnist.py --corr 0.995 --alpha 1500  
python train_biased_mnist.py --corr 0.997 --alpha 2000  
python train_biased_mnist.py --corr 0.999 --alpha 10000  

python train_utk_face.py --task race  
python train_utk_face.py --task age  

python train_celeba.py --task blonde --alpha 30000  
python train_celeba.py --task makeup --alpha 20000  

python get_imagenet_bias_features.py  
python train_imagenet.py  

If you find this code useful in your research, please consider citing:

@article{sarridis2024flac,  
  title={Flac: Fairness-aware representation learning by suppressing attribute-class associations},  
  author={Sarridis, Ioannis and Koutlis, Christos and Papadopoulos, Symeon and Diou, Christos},  
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},  
  year={2024},  
  publisher={IEEE}  
}