Table 6 Scores for the high confidence region for matching pairs (\(\textbf{z}_m\)) and non-matching pairs (\(\textbf{z}_n\)) together with the maximum accuracy obtained for LFW, YTF, CALFW, CPLFW, CFP, and AgeDB datasets
From: Cancelable templates for secure face verification based on deep learning and random projections
Dataset | \((\mu -3\sigma \le \textbf{z}_m \le \mu +3\sigma )\) | \((\mu -3\sigma \le \textbf{z}_n \le \mu +3\sigma )\) | Max accuracy |
|---|---|---|---|
LFW | 99.33% | 95.60% | 99.82 |
YTF | 95.37% | 86.23% | 98.16 |
CALFW | 91.67% | 95.80% | 97.10 |
CPLFW | 85.27% | 84.30% | 95.85 |
CFP | 98.40% | 94.73% | 99.43 |
AgeDB | 86.27% | 83.03% | 96.24 |