A new biometric research information or data set is now accessible from the National Institute of Standards and Technology. The data set ranges from iris scans to facial photographs, to fingerprints. The new data set is available strictly for research purposes only and does not offer any identification information. The design of the data set primarily concentrates on testing systems, which are able to verify the identity of a person before giving access. It can identify a person sitting in another room or in a different country. There are only a handful of resources available today that help developer in evaluating the performance of the software algorithms. These software algorithms are at the core of these biometric identification systems. Moreover, the NIST data sets will now help in bridging the gap between the systems.
According to Greg Fiumara, a NIST computer scientist, this all points to reproducible research. Furthermore, he adds that the data sets will aid anyone who shows interest in testing and solving error rates in biometric identification systems.
What are the Special Data Bases?
The files – now available on the website of NIST – have three distinct ‘Special Databases (SDs)’. They are called SD 300, SD 301, and SD 302. Moreover, these special databases represent the first of an ever-growing collection of biometric resources.
While these three databases consist of a broad range of data, two of these sets contain data collected at the IARPA funded competition – Nail to Nail Fingerprint Challenge. Moreover, NIST helped in creating a design and execution of this competition.
SD 301, one of the three new special databases, contains NIST’s first-ever multimodal dataset. Multimodal data represents different biometric markers of an individual. In this particular scenario, it represents iris, fingerprints, and face scans of an individual. They all link together to form an identification metric for the biometric systems.
The biometric system then uses different combinations of these metrics to identify an individual.