Panasonic announces deep learning-powered facial recognition software

Panasonic Corporation has announced that it will launch server software for facial recognition powered by deep learning technology in July 2018.

Face Recognition Server Software WV-ASF950 and Face Registration Expansion Kit WV-ASFE951W feature a core engine with the world’s highest performance, according to the company, and can identify faces that pose difficulty for conventional technology. These include faces captured at an angle up to 45 degrees to the left or right, or up to 30 degrees above or below, as well as those partially hidden by sunglasses. The company also plans to add a function for faces partially covered by a surgical mask. The Face Registration Expansion Kit is sold separately, and enables up to 30,000 faces at large-scale facilities.

Panasonic’s performance claim is based on testing in April, 2017, against the NIST IJB-A face challenge, according to the announcement. The company also said its deep learning technology, which was jointly developed with the National University of Singapore, reduced the rejection rate to 20 percent when false accept rate was set to 0.01 with the IJB-A face image dataset, an improvement of 500 percent over conventional systems.

The software also features “iA (intelligent automatic) mode,” which automatically adjusts camera settings to optimize images for facial recognition. It can be used with Panasonic’s i-PRO Extreme series network cameras and the “Best Shot License Key” which is bundled with the software to send only the best images for facial recognition to the server for matching. The iA function also enables image analysis to be performed on-camera to reduce network bandwidth needs and reduce overall system cost. Panasonic says system costs can be reduced by 40 to 50 percent using Best Shot.

The software will be launched initially outside of Japan, with the Japanese release slated for the following month.

Panasonic face recognition contracts include a recently announced selection to streamline immigration procedures at a major Japanese airport.

Article Topics

biometric software  |  deep learning  |  facial recognition  |  Panasonic