VeriEye SDK
Iris identification for stand-alone and Web solutions
VeriEye iris identification technology is designed for biometric systems developers and integrators. The technology includes many proprietary solutions that enable robust iris enrollment under various conditions and fast iris matching in 1-to-1 and 1-to-many modes.
Available as a software development kit that allows development of stand-alone and Web-based solutions on Microsoft Windows, Linux, Mac OS X, iOS and Android platforms.
Reliability Tests
We present the testing results to show VeriEye 10.0 template matching algorithm reliability evaluations. Iris images from several standard datasets were used for testing, thus the testing results can be compared with testing results of other algorithms. All datasets contained iris images with 640 x 480 pixels size.
| Iris image datasets used for VeriEye 10.0 algorithm testing | ||||
|---|---|---|---|---|
| ICE2005 Exp1 (1) | ND-IRIS-0405 (2) | IRISDB1600 (3) | MobileIrisV10 (4) | |
| Image count | 1,425 | 64,980 | 24,361 | 3,290 |
| Subject count | 124 | 356 | 624 | 70 |
| Unique iris count | 124 | 712 | 1231 | 135 |
| Session count | 1 - 31 | 4 - 291 | 1 - 40 | 6 - 42 |
Notes:
- The ICE2005 dataset was collected by the National Institute of Standards and Technology (NIST). Near-infrared spectrum equipment was used for iris capture. ICE2005 Exp1 is a subset, which contains right iris images.
- The ND-IRIS-0405 was collected by the University of Notre Dame. Near-infrared spectrum equipment was used for iris capture.
- The IRISDB1600 was collected by the University of Bath. Near-infrared spectrum equipment was used for iris capture. The full IRISDB1600 dataset contains 31,997 images (image size 1280x960 pixels), which represented 799 unique persons and 1,598 unique irises. A subset used in this test was preprocessed similar to NIST IREX experiments – the images were downsampled to 640x480 via 2x2 neighborhood averaging, and all images containing irises with diameters larger than 340 pixels were removed.
- The MobileIrisV10 dataset was collected by the Warsaw University of Technology. The iris image collection was performed using regular, visible light spectrum camera built-in into Apple iPhone 5S smartphone. Colored images were collected with the camera. The images were resized to 640x480 pixels and converted to grayscale. See the scientific paper for more details.
Two tests were performed with each dataset:
- Test 1 maximized matching accuracy. VeriEye 10.0 algorithm reliability in this test is shown as blue curves on the ROC charts.
- Test 2 maximized matching speed. VeriEye 10.0 algorithm reliability in this test is shown as red curves on the ROC charts.
The iris rotation tolerance was set to ±15° in all tests.
Receiver operation characteristic (ROC) curves are usually used to demonstrate the recognition quality of an algorithm. ROC curves show the dependence of false rejection rate (FRR) on the false acceptance rate (FAR).
ICE2005 Exp1
Click to zoom |
ND-IRIS-0405
Click to zoom |
Bath IRISDB1600
Click to zoom |
MobileIrisV10
Click to zoom |
| VeriEye 10.0 algorithm reliability testing results | ||||||||
|---|---|---|---|---|---|---|---|---|
| ICE2005 Exp1 | ND-IRIS-0405 | IRISDB1600 | MobileIrisV10 | |||||
| Test 1 | Test 2 | Test 1 | Test 2 | Test 1 | Test 2 | Test 1 | Test 2 | |
| FRR at 0.01 % FAR | 0.0123 % | 0.0942 % | 1.0420 % | 1.2590 % | 0.0373 % | 0.0496 % | 0.0527 % | 1.1490 % |
| FRR at 0.001 % FAR | 0.0246 % | 0.1064 % | 1.3240 % | 1.5300 % | 0.0453 % | 0.0587 % | 0.0527 % | 1.3380 % |
