English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Generative Adversarial Networks (GANs) advance face synthesis throughlearning the underlying distribution of observed data. Despite the high-qualitygenerated faces, some minority groups can be rarely generated from the trainedmodels due to a biased image generation process. To study the issue, we firstconduct an empirical study on a pre-trained face synthesis model. We observethat after training the GAN model not only carries the biases in the trainingdata but also amplifies them to some degree in the image generation process. Tofurther improve the fairness of image generation, we propose an interpretablebaseline method to balance the output facial attributes without retraining. Theproposed method shifts the interpretable semantic distribution in the latentspace for a more balanced image generation while preserving the samplediversity. Besides producing more balanced data regarding a particularattribute (e.g., race, gender, etc.), our method is generalizable to handlemore than one attribute at a time and synthesize samples of fine-grainedsubgroups. We further show the positive applicability of the balanced datasampled from GANs to quantify the biases in other face recognition systems,like commercial face attribute classifiers and face super-resolutionalgorithms.

Improving the Fairness of Deep Generative Models without Retraining