The data used in IREX comes from publicly available datasets and from operational data provided by various government and commercial entities. The publicly available data sets may be procured from the original sources; in general, NIST does not have license to redistribute any of the public datasets that are used in IREX. The operational data has been provided by the data owners under the condition that it will not be redistributed by NIST. Hence, in general, the IREX program cannot provide datasets to the research community. A list of the datasets used in the IREX effort can be found here .
IREX IX continues along the path of IREX III & IV as an evaluation of iris recognition algorithms for use in large-scale deployments. IREX IX will explore:
- Verification: One-to-one matching
- Identification: One-to-many matching
- Effects of illumination wavelength: visible through the near infrared, both within and between wavelengths
- Effects of off-angle images: such as might be expected from captures on handheld mobile devices
- Forensic applications: images from cameras not designed for iris recognition
As before, the primary evaluations will be carried out on sequestered data from US government sources – this assures a level playing field for all. Secondary evaluations on publicly available datasets will also be carried out – this aids developers by enabling reproduction of evaluations.
NIST expects to coordinate IREX VIII to support implementation of the iris recognition option extended under NIST Special Publication 800-76-2. This activity will constitute a laboratory evaluation of iris recognition algorithms capable of producing and consuming conformant instances of ISO/IEC 19794-6:2011. The schedule for this activity is to be determined.
IREX VII defines a framework for communication and interaction between components in an iris recognition system by introducing layers of abstraction that isolate underlying vendor-specific implementation details from application programs in the system. This enables system development that is more flexible, extensible, and modifiable, as well as easier reuse of system components. IREX VII is not intended as a replacement for Biometric Web Services or Bio-API; it is a complement to both that can thought of as the first level above hardware or algorithm in a protocol stack. IREX VII documents are available on the IREX VII homepage.
In response to studies reporting an aging effect in iris recognition, NIST initiated the IREX VI project to independently assess the stability of the iris for long term recognition, and to track other research in the area. The IREX VI homepage includes NIST research on longitudinal analysis of iris recognition systems.
IREX V provides best practice recommendations and guidelines for the proper collection and handling of iris images. IREX V documents are available on the IREX V homepage
IREX IV continues along the path of IREX III as an evaluation of one-to-many iris recognition for large-scale applications. IREX IV was initiated to serve two purposes: 1) to explore the potential for using cost functions for application-specific algorithm optimization; and 2) to support the ISO/IEC 19794-6 standard by defining a compression profile for the compact representation of iris images.
NIST is happy to announce the availability of first part of the IREX IV evaluation report, NIST Interagency Report 7949. The second and final part, NIST Interagency Report 7978,focuses on compression profiles for iris image compression
NIST concluded the IREX III evaluation of one-to-many iris algorithms in April 2012. IREX III was the first independent and public test of iris identification search technologies. It used millions of images to validate results published in the academic literature that iris is a very powerful biometric. The final report and a chronology of the test are given on the IREX III home page:
IREXII IQCE aims to evaluate the effectiveness of image quality assessment algorithms (IQAAs) that produce a scalar overall image quality in predicting the recognition accuracy of particular comparison algorithms (from the supplier of the IQAA), and of other algorithms. Furthermore, per the IREX I result that quality scores are not immediately interoperable, IQCE will establish a score calibration procedure for IQAAs.
IREX I was being conducted to address two issues: First is how far compression can be pushed before the accuracy of leading commercial matching algorithms begins to degrade. Second, the accuracy of iris images in compact polar form when it is prepared by one supplier and verified later by another. The program supports identity management applications where compact size and interoperability are of primary concern. These include federated identity credential or network-based applications.