Closed Athuli7 closed 4 years ago
I'm using the linked fingerprint database.
Which dataset is this? Judging by filenames, this is one of the Neurotech datasets that contain 400-500 fingerprints that form 160K-250K pairs, most of them non-matching. Finding 14 false matches in so many pairs is actually very close to 0.01% FMR.
Nevertheless, the relationship between threshold and FMR is approximate. If your dataset contains large prints with lots of minutiae, you should probably raise the threshold a bit.
Main.java
User.java
Problem section of console output
The threshold of 40 is triggering false positives. This is nowhere close to 0.01 FAR. Or am I doing something wrong?