Adaptive Calibration for Fair and Performant Facial Recognition
cs.CV, cs.AI
Submitted: 2026-06-03
Updated: 2026-09-08
Comments: 36 pages, 2 figures. Revised implementation details and discussion; added component ablations, LFW validation, and robustness analyses
License: http://creativecommons.org/licenses/by/4.0/
The gist: We introduce Adaptive Calibration (AC), a novel calibration strategy for facial recognition that maps cosine similarity between normalized embeddings to well-calibrated probabilities.
Terminology
Abstract
We introduce Adaptive Calibration (AC), a novel calibration strategy for facial recognition that maps cosine similarity between normalized embeddings to well-calibrated probabilities. By incorporating local context into calibration, Adaptive Calibration corrects for a fundamental mismatch in cosine similarity, whereby the same distance can correspond to different match probabilities in different embedding regions. Our approach improves both overall performance and results in a fairer calibration without requiring demographic metadata. Our approach often improves worst-group ranking and probability quality across a variety of pretrained models and standard benchmarks, with gains in low-FPR verification performance depending on the setting. AC provides a practical solution for equitable facial recognition, without requiring demographic group annotations, and while improving overall performance. Our method provides continuous, region-specific calibration, and we examine per-group performance to assess "leveling down" where fairness comes at the cost of degraded performance for some groups.
Sources
- A Deep Dive into Dataset Imbalance and Bias in Face Identification
- Mitigating Gender Bias in Face Recognition Using the von Mises-Fisher Mixture Model
- OxonFair: A Flexible Toolkit for Algorithmic Fairness
- ArcFace: Additive Angular Margin Loss for Deep Face Recognition
- PASS: Protected Attribute Suppression System for Mitigating Bias in Face Recognition
- Towards Gender-Neutral Face Descriptors for Mitigating Bias in Face Recognition
- Deep Imbalanced Learning for Face Recognition and Attribute Prediction
- Review of Demographic Fairness in Face Recognition
- Score Normalization for Demographic Fairness in Face Recognition
- SphereFace: Deep Hypersphere Embedding for Face Recognition
- The Unfairness of Fair Machine Learning: Levelling down and strict egalitarianism by default
- On Fairness and Calibration
- Face Recognition: Too Bias, or Not Too Bias?
- FairCal: Fairness Calibration for Face Verification
- FaceNet: A Unified Embedding for Face Recognition and Clustering
- Face Recognition: A Novel Multi-Level Taxonomy based Survey
- FRAPPE: A Group Fairness Framework for Post-Processing Everything
- CosFace: Large Margin Cosine Loss for Deep Face Recognition
- Racial Faces in-the-Wild: Reducing Racial Bias by Information Maximization Adaptation Network
Related papers
- Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
- AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
- Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
- MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
- TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
- A Survey on Efficient Vision-Language-Action Models