Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

FairCal: Fairness Calibration for Face Verification

About

Despite being widely used, face recognition models suffer from bias: the probability of a false positive (incorrect face match) strongly depends on sensitive attributes such as the ethnicity of the face. As a result, these models can disproportionately and negatively impact minority groups, particularly when used by law enforcement. The majority of bias reduction methods have several drawbacks: they use an end-to-end retraining approach, may not be feasible due to privacy issues, and often reduce accuracy. An alternative approach is post-processing methods that build fairer decision classifiers using the features of pre-trained models, thus avoiding the cost of retraining. However, they still have drawbacks: they reduce accuracy (AGENDA, PASS, FTC), or require retuning for different false positive rates (FSN). In this work, we introduce the Fairness Calibration (FairCal) method, a post-training approach that simultaneously: (i) increases model accuracy (improving the state-of-the-art), (ii) produces fairly-calibrated probabilities, (iii) significantly reduces the gap in the false positive rates, (iv) does not require knowledge of the sensitive attribute, and (v) does not require retraining, training an additional model, or retuning. We apply it to the task of Face Verification, and obtain state-of-the-art results with all the above advantages.

Tiago Salvador, Stephanie Cairns, Vikram Voleti, Noah Marshall, Adam Oberman• 2021

Related benchmarks

TaskDatasetResultRank
Face VerificationBFW
TPR @ FPR 0.1%93.94
138
Face VerificationLFW
AUROC99.32
67
CalibrationBFW
Worst-Group Brier Score0.042
66
CalibrationLFW
Worst-group Brier score0.063
66
Face VerificationLFW
Min-Group AUROC (%)97.6
66
Face VerificationRFW
Min-Group AUROC98.83
66
CalibrationRFW
Worst-group Brier Score0.056
66
Face VerificationLFW
EO Gap (0.1%)13.82
43
Face VerificationRFW
TMR @ FMR 1e-30.742
36
Face VerificationBFW headline settings
AUROC98.4
33
Showing 10 of 27 rows

Other info

Follow for update