Face comparison¶
Face comparison is the process of comparing a face template against a previously stored reference template.
The FaceMatcher module provides methods for comparing two face templates or two face template lists. The methods return a comparison score, from 0 to 65535 representing the degree of similarity between the two templates.
Note
A FaceTemplateList should contain templates from a single user.
Decision threshold¶
The decision is the result of the comparison between the biometric comparison similarity score of given biometric data and a predefined threshold to achieve the desired security level. If the score is not less than the predefined minimal matching score threshold, the biometric verification is successful.
A decision threshold applied on the algorithm’s output score determines the operational FMR (False Match Rate), i.e. the probability of a system to falsely accept faces (impostors). Since FMR and FNMR (False Non-Match Rate) are in inverse proportion to each other, FNMR will increase with higher decision thresholds.
The decision threshold is defined according to the following table:
False Match Rate |
Decision threshold |
|---|---|
1.0% |
2000 |
0.1% |
3000 |
0.01% |
4000 |
0.001% |
5000 |
0.0001% |
6000 |
0.00001% |
7000 |
0.000001% |
8000 |
0.000001% |
9000 |
0.0000001% |
10000 |
Hint
To obtain a value between 0 and 100, we recommend applying the following formula to the matching score (S):
\(S' = MIN(100, S * 8 / 250)\)
Example¶
The example below demonstrates how to compare two faces:
# initialize a FaceMatcher
face_matcher = FaceMatcher()
# Compare the templates
score = face_matcher.compare_templates(reference, probe)
print(f"Score = {score}")
print(f"Similarity = {min(100, score * 8 / 250):0.2f}%")
# Dispose all resources
del face_matcher