TECH5 Posts Fastest Search Time in NIST Iris Developer Table

TECH5 has recorded a search time of 0.088 seconds against a gallery of 500,000 identities in the National Institute of Standards and Technology Iris Exchange 10 Identification Track, the lowest search time currently reported in the evaluation’s two-eye developer table.

The submission, tech5_009, returns a template of 6,237 bytes and a false negative identification rate of 0.0067 at a false positive identification rate of 0.01. The search time carries a margin of 0.002 seconds and the template size a margin of 99 bytes. The company frames the result as a balance of accuracy, speed and storage rather than a win on accuracy alone.

The comparison TECH5 draws is with the five algorithms currently ranked highest for two-eye accuracy in the same NIST table. Those entries record search times between 4.6 and 12 seconds and templates between 18,280 and 130,879 bytes. On those published figures TECH5 puts its engine at roughly 52 to 136 times faster in the identification search, using approximately 66 to 95 percent less template storage. Measured against the algorithm ranked first for two-eye accuracy, which runs on the same newer NIST test platform, the published figures show 4.6 seconds against 0.088 seconds, and 19,349 bytes against 6,237 bytes.

Template size governs how much biometric data a system must store, load, cache and replicate as galleries grow. Extrapolating the published template sizes to a hypothetical gallery of 100 million people, TECH5 puts its own raw template payload at roughly 624 gigabytes against roughly 1.83 to 13.09 terabytes for the five highest-accuracy entries. The company notes those figures exclude database structures, indexes, redundancy and backups.

TECH5 also cautions that matcher timings are not end-to-end passenger processing times, since image capture, networking, application logic and database access all add latency in a deployed system.

The company has used NIST evaluations to position its engines for population-scale work before, reporting fingerprint 1:N results at national ID scale, and its infrastructure underpins national programmes including Papua New Guinea’s national digital ID and wallet. Iris continues to draw operational interest, with US Customs and Border Protection expanding its use of the modality.

Sources: TECH5

–

By the ID Tech Editorial Team