Biometric accuracy with masks added as criteria for Indian public face recognition system

The biometric facial recognition system being stood up by India’s National Crime Records Bureau (NCRB) will have to be capable of identifying masked faces, so companies bidding for the contract will be asked to showcase how well they work on occluded faces, MediaNama reports.

The evaluation process for the Automated Facial Recognition System (AFRS) will now consider effectiveness at biometric identity-matching among the decision criteria, according to a document acquired by the media outlet, which responds to vendor questions. Twenty vendors posed a total of 333 questions or requests for clarification.

Other points clarified by the document include the need for the winning vendor to develop APIs for integrating the AFRS with other databases, like the CCTNS and NAFIS, in consultation with the agencies that operate them, and accuracy standards of a true positive match among the top two results for frontal images, and 9 hits out of 10 among the top 10 results for faces captured at an angle. The system will not have to integrate with a video management system.

Several of the companies planning to bid on the contract, which has already been delayed ten times, have tuned their algorithms to identify people wearing masks. Among the companies bidding on the contract, NEC has already announced an upgrade of its facial recognition engine for matching masked faces, while AnyVision recently declared its technology works with masked faces.

NIST testing has shown algorithms not specifically trained to recognize masked faces tend to struggle with them.

Madurai police launch face recognition app

Police in Madurai, Tamil Nadu have implemented an app-based biometric facial recognition service, known as FACETAGR, to identify wanted individuals, Businessworld reports.

The Android application calls back to a web-based system, according to the report. Madurai’s police commissioner also said the force is in talks with a neighbouring district to integrate their databases of convicted criminals and suspects.

The app works with images only 50 kb in size, returning matches with confidence scores in less than a second, even with slow internet connections, Businessworld writes.

The app can be used only be police officers who are given non-transferable one-time access codes.

Kerala considers biometric migrant checks

Police in Kerala are considering requesting access to a biometric database of migrant laborers held by the Labour Department as part of an attempt to identify criminals and terrorists claiming to be migrants, according to The New Indian Express.

The National Investigation Agency (NIA) recently arrested three suspected members of Al-Queda in the region, prompting state police to increase surveillance measures. More than 500,000 migrant laborers were registered by the Labour Department for migrant health insurance program Awaaz, and there estimated to be as many as 2.5 million migrant workers in the state.

Previous attempts by police to collect migrant biometrics were resisted as an infringement of privacy rights, according to the report.

Article Topics

biometric database  |  biometrics  |  criminal ID  |  facial recognition  |  India  |  mask detection  |  mobile app  |  police  |  tender