Fixing Class C FNIR Values for Recent Algorithms
2 views
Skip to first unread message
Gregory Fiumara
May 21, 2026, 11:10:22 AMMay 21
to Friction Ridge Image and Features Technology Evaluations
There was a discrepancy noticed between columns for Class C FNIR values on our table of results and in individual participant reports. This has been corrected.
Reason
Around mid-March, we were looking into a suspicion of multiple correct mates for mated Class C searches being present in the databases (i.e., the same 10 fingerprints under more than one subject identifier). In a trial, we removed these searches from the test and re-ran analysis on a small number of algorithms to see the effect. Unfortunately, this trial analysis crept into production as we introduced the results table and published new reports this month. We have reverted the change until we can review the images in question further.
Summary of Changes
Incorrect data appeared in Class C columns in the results table for the following participants and has been corrected:
My apologies for the error and thanks for your understanding!
-Greg
Reason
Around mid-March, we were looking into a suspicion of multiple correct mates for mated Class C searches being present in the databases (i.e., the same 10 fingerprints under more than one subject identifier). In a trial, we removed these searches from the test and re-ran analysis on a small number of algorithms to see the effect. Unfortunately, this trial analysis crept into production as we introduced the results table and published new reports this month. We have reverted the change until we can review the images in question further.
Summary of Changes
Incorrect data appeared in Class C columns in the results table for the following participants and has been corrected:
- hisign+000C
- IDEMIA+0001
- Neurotechnology+0104
- Neurotechnology+0106
- roc+0003
My apologies for the error and thanks for your understanding!
-Greg
Reply all
Reply to author
Forward
0 new messages