Anatomizing Bias in Facial Analysis

Singh, Rajesh; Majumdar, Puspita; Mittal, Surbhi; Vatsa, Mayank

doi:10.48550/arxiv.2112.06522

Cited by 1 publication

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…2019 Face Recognition Vendor test documents lower female accuracy rates across a broad range of algorithms and datasets 4 . Similarly, lower accuracy rates for females have been obtained for various in-house deep learningbased face recognition systems [5,4,30]. The cause and effect analysis suggests gendered hairstyles resulting in facial occlusion, make-up, and inherent lower variability between different female faces over males to be the factors contributing to lower performance for females [4,5].…”

Section: Gendered Differences In Facial Analyticsmentioning

confidence: 97%

“…This draws attention to fairness and bias in AI-based facial analytics where unintended consequences from biased systems call for a thorough examination of the datasets and models [18,5,17,8,4]. Most of the published research in this domain suggests low performance for women, and dark-skinned people for facial attribute-based classification systems such as gender and age [8,17,30,24], and face recognition [5,4]. As biased datasets produce biased models, many of the efforts have been focused on developing gender and race-balanced datasets for various facial-analysis based applications.…”

Section: Celebdf Distributionmentioning

confidence: 99%

See 1 more Smart Citation

GBDF: Gender Balanced DeepFake Dataset Towards Fair DeepFake Detection

Nadimpalli¹,

Rattani²

2022

Preprint

View full text Add to dashboard Cite

Facial forgery by deepfakes has raised severe societal concerns. Several solutions have been proposed by the vision community to effectively combat the misinformation on the internet via automated deepfake detection systems. Recent studies have demonstrated that facial analysis-based deep learning models can discriminate based on protected attributes. For the commercial adoption and massive roll-out of the deepfake detection technology, it is vital to evaluate and understand the fairness (the absence of any prejudice or favoritism) of deepfake detectors across demographic variations such as gender and race. As the performance differential of deepfake detectors between demographic subgroups would impact millions of people of the deprived sub-group. This paper aims to evaluate the fairness of the deepfake detectors across males and females. However, existing deepfake datasets are not annotated with demographic labels to facilitate fairness analysis. To this aim, we manually annotated existing popular deepfake datasets with gender labels and evaluated the performance differential of current deepfake detectors across gender. Our analysis on the gender-labeled version of the datasets suggests (a) current deepfake datasets have skewed distribution across gender, and (b) commonly adopted deepfake detectors obtain unequal performance across gender with mostly males outperforming females. Finally, we contributed a gender-balanced and annotated deepfake dataset, GBDF, to mitigate the performance differential and to promote research and development towards fairness-aware deep fake detectors. The GBDF dataset is publicly available at: https://github.com/aakash4305/GBDF

show abstract