Final Results

Three teams submitted their methods: Aurora, HH (3 methods) and NTNU (3 periocular methods and 3 iris methods).

Table 1. Results of the periocular methods submitted (error rate in %)
 Team
Rank 
GF4
GFRR@GFAR=0.01% 
GF3
GFRR@GFAR=0.1%
Processing time (Enrol) Avg. (sec)  Processing time (Match) Avg. (sec)
 HH1
1
0.94
 0.5418.05
0.70
HH2
2
22.91
 15.9914.40
0.05
 HH3
3
26.32
 19.0814.39
0.05
 NTNU2
4
31.25
 28.906.03
5.18
 NTNU3
 5 
36.84
 34.566.03
5.16
NTNU1
 6 40.3538.02
 6.03 5.15
 Aurora*7
90.00
89.56
13.47
8.39
* The method submitted by team Aurora was not compliant to our guidelines.

Table 2. Results of the iris methods submitted (error rate in %)
 Team
Rank 
GF4
GFRR@GFAR=0.01% 
GF3
GFRR@GFAR=0.1%
Processing time (Enrol) Avg. (sec)  Processing time (Match) Avg. (sec)
 NTNU2
1
91.77
5.15
118.68
5.15
 NTNU3
 2
95.83
5.98
118.685.15
NTNU1
3 97.605.38
 118.68 5.14



1. For the final submission evalutation we used a test set with 80 users for periocular and 160 for iris.

2. All possible cross-sensor intra-class comparisons are implemented to evaluate the false non-match rate (FNMR).

3.Three samples were selected from each class for each sensor to evaluate the false match rate (FMR).

4.
The metric GFRR@ GFAR=0.01% was used as the performance indicator to rank the submitted algorithms. GF3 is also reported for a better characterization of the algorithm's performance.

5. The method submitted by team Aurora was not compliant to our guidelines. The results published here are not directly comparable to the others.

Comments