Adaptive Calibration for Fair and Performant Facial Recognition
Quick summary
arXiv:2606.04469v2 Announce Type: replace-cross 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. O
Key takeaways
- arXiv:2606.04469v2 Announce Type: replace-cross 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.
Why it matters
The importance of “Adaptive Calibration for Fair and Performant Facial Recognition” will be measured by what changes in practice. User behavior, access conditions, verifiable performance and responsible-use outcomes are the signals worth following.
Member comments