Improving the Fairness of Deep Generative Models without Retraining

Abstract

Generative Adversarial Networks (GANs) advance face synthesis through learning the underlying distribution of observed data. Despite the high-quality generated faces, some minority groups can be rarely generated from the trained models due to a biased image generation process. To study the issue, we first conduct an empirical study on a pre-trained face synthesis model. We observe that after training the GAN model not only carries the biases in the training data but also amplifies them to some degree in the image generation process. To further improve the fairness of image generation, we propose an interpretable baseline method to balance the output facial attributes without retraining. The proposed method shifts the interpretable semantic distribution in the latent space for a more balanced image generation while preserving the sample diversity. Besides producing more balanced data regarding a particular attribute (e.g., race, gender, etc.), our method is generalizable to handle more than one attribute at a time and synthesize samples of fine-grained subgroups. We further show the positive applicability of the balanced data sampled from GANs to quantify the biases in other face recognition systems, like commercial face attribute classifiers and face super-resolution algorithms.

Project page: https://genforce.github.io/fairgen/

References in corpus (11)

• GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Equality of Opportunity in Supervised Learning
• The Measure and Mismeasure of Fairness
• Fairness in Machine Learning
• Learning Non-Discriminatory Predictors
• Flexibly Fair Representation Learning by Disentanglement
• InclusiveFaceNet: Improving Face Attribute Detection with Race and Gender Diversity
• Bias and Generalization in Deep Generative Models: An Empirical Study
• A statistical framework for fair predictive algorithms
• Imperfect ImaGANation: Implications of GANs Exacerbating Biases on Facial Data Augmentation and Snapchat Selfie Lenses

Cited by in corpus (5)

• Bias-Free FedGAN: A Federated Approach to Generate Bias-Free Datasets
• Generative Models as a Data Source for Multiview Representation Learning
• Measuring Fairness in Generative Models
• Fairness in generative modeling
• MaGNET: Uniform Sampling from Deep Generative Network Manifolds Without Retraining