This September, the National Institution of Standards and Technology (NIST) has welcomed the quality of an algorithm made by an Indonesian based AI tech company, Nodeflux. They put Nodeflux in the 25th position from a total of 90 AI tech companies competing in the Face Recognition Vendor Test (FRVT), with those from China and Russia in the same category.
NIST is a standardize institution and science & technology laboratory under the US Trading Department. It was to create a greater battle in tech development worldwide. One of the programs is FRVT as the fine benchmark of face recognition based on its algorithm.
“NIST benchmark is very helpful for vendors using face recognition to gain an assessment of its technology. Nodeflux, in the 25th place for the Wild 1E-4 dataset category worldwide, is making us very proud for bringing the Indonesian’ tech development to the global competition,” Nodeflux‘ Co-Founder and CEO, Meidy Fitranto said.
The assessment has three categories, Visa, Mugshot, and Wild Dataset by evaluating the identification performance through scenarios, ethnics, gender, and ages. The dataset testing works on some scenarios on field, such as territorial borders, ID access, and city safety.
Face recognition was done with a comparison concept between input and reference images consist of two types, 1:1 (one to one), comparison of 1 input image with 1 reference image and 1:N (one to many), comparison of 1 input image with various images of all sides.
“In order to get into the Wild 1E-4 dataset category for the 1:1 type, our engineers are passing through the FRTV trial program assessment. In competition with AI giants from China and US is one challenge. We keep running trial to get the most accurate performance, up to this position. Furthermore, there will be improvements for the current technology,” Nodeflux’ Co-Founder and CTO, Faris Rahman.
–
Original article is in Indonesian, translated by Kristin Siagian