4 FpVTE 2012—Class A

Results in this section involve plain impression index fingers. Probes containing left index, right index, and both index fingers were searched against an enrollment database consisting of both index fingers. Individual index finger probes were searched against an enrollment database of 100 000 subjects, while probes containing both index fingers are searched against an enrollment database of 1 600 000 subjects.

The datasets in this section are equivalent to those used in NIST FpVTE 2012 (Class A). Detailed information about FpVTE 2012 can be found in NIST IR 8034.

Notes:

  • No examiner extended feature set data was provided with the images.

4.1 Template Generation

The approximate total number of index finger records that underwent template generation along with a tally of records that failed to process are shown in Table 4.1.

Table 4.1: Summary of template generation for FpVTE 2012—Class A.
Image Contents Template Type Failure to Extract ≈Total
Left Index (Plain) Probe 0 30 000
Right Index (Plain) 0 30 000
Left + Right Index (Plain) 0 30 000
Reference 0 100 000
0 1 600 000

4.2 Both Index Fingers

Probe templates with features extracted from 30 000 pairs of index fingers were searched against an enrollment database of 1 600 000 subjects. Each probe template was generated by a single function call providing tech5impl+0005 two separate images (i.e., left index and right index). Likewise, each subject template incorporated into the database was generated by a single function call providing tech5impl+0005 two separate images (i.e., left index and right index). Approximately one-third of the probes had a corresponding mate in the enrollment database. Each subject in the enrollment database had only left and right index fingers represented.

4.2.1 DET

The DET plots in Figure 4.1 show the tradeoff of errors of tech5impl+0005 when searching pairs of index fingers from FpVTE 2012—Class A against enrollment database of 1 600 000 subjects where, for approximately one-third of the probes, a single mated identity consisting of left and right index fingers was present. Tabular versions of FNIR at select FPIR can be viewed in Table 4.2.

DET when searching both index fingers against an enrollment database of both index fingers.

Figure 4.1: DET when searching both index fingers against an enrollment database of both index fingers.

Table 4.2: FNIR values from the DET plotted in Figure 4.1.
Probe Content FPIR  ≤ 0.001 FPIR  ≤ 0.005 FPIR  ≤ 0.01
Left + Right Index (Plain) 0.001 0.0007 0.0006
Table 4.3: Similarity score thresholds from the DET plotted in Figure 4.1.
Probe Content FPIR  ≤ 0.001 FPIR  ≤ 0.005 FPIR  ≤ 0.01
Left + Right Index (Plain) 0.2719 0.2491 0.2429

4.2.2 CMC

The CMC plots in Figure 4.2 show the FNIR of tech5impl+0005 when searching pairs of index fingers from FpVTE 2012—Class A against enrollment database of 1 600 000 subjects where, for approximately one-third of the probes, a single mated identity consisting of left and right index fingers was present. Tabular versions of FNIR at select ranks can be viewed in Table 4.4.

CMC when searching both index fingers against an enrollment database of both index fingers.

Figure 4.2: CMC when searching both index fingers against an enrollment database of both index fingers.

Table 4.4: FNIR values from CMC plotted in Figure 4.2.
Probe Content Rank 1 Rank  ≤ 2 Rank  ≤ 5 Rank  ≤ 10 Rank  ≤ 50 Rank  ≤ 100
Left + Right Index (Plain) 0.0004 0.0003 0.0003 0.0003 0.0002 0.0002