Sampling Strategies for Mitigating Bias in Face Synthesis Methods
This addresses bias in synthetic face generation for media and training datasets, but it is incremental as it builds on existing methods.
The paper tackles bias in face synthesis by analyzing StyleGAN2 trained on Flickr Faces HQ, revealing under-representation of very young, very old, and female faces, and proposes two latent space sampling strategies that reduce bias and improve uniformity across image quality levels.
Synthetically generated images can be used to create media content or to complement datasets for training image analysis models. Several methods have recently been proposed for the synthesis of high-fidelity face images; however, the potential biases introduced by such methods have not been sufficiently addressed. This paper examines the bias introduced by the widely popular StyleGAN2 generative model trained on the Flickr Faces HQ dataset and proposes two sampling strategies to balance the representation of selected attributes in the generated face images. We focus on two protected attributes, gender and age, and reveal that biases arise in the distribution of randomly sampled images against very young and very old age groups, as well as against female faces. These biases are also assessed for different image quality levels based on the GIQA score. To mitigate bias, we propose two alternative methods for sampling on selected lines or spheres of the latent space to increase the number of generated samples from the under-represented classes. The experimental results show a decrease in bias against underrepresented groups and a more uniform distribution of the protected features at different levels of image quality.