Marginal debiased network for fair visual recognition

Mei Wang, Weihong Deng, Jiani Hu, Sen Su

Abstract: Highlights•We propose a novel marginal debiased network for fair visual recognition.•Margin penalty is used to emphasize bias-conflicting data and mitigate spurious correlation.•We develop a meta learning framework to automatically learn the optimal margins.•Extensive experiments demonstrate the superiority of our method.