Facetagr offers facial recognition to ID missing Nepal flood victims

Nepalese authorities have adopted an AI-powered facial recognition system to complement DNA analysis efforts aimed at accelerating the identification of dead persons found after devastating floods hit parts of the country last month.

U.S. company NowtrKit has partnered with India-based facial recognition developer  FaceTagr to match faces captured in photos and videos with people who have been declared missing, according to local media reports. FaceTagr has submitted multiple biometric algorithms to NIST’s FRTE 1:1, the latest back in 2022. The company has worked with the Bihar State Election Commission on local election security and with police in Tamil Nadu on suspect identification software.

Vijay Gnanadesikan, Facetagr co-founder and CEO, calls it “a humanitarian, not-for-profit initiative.”

The process to identify recovered bodies has stalled over reported limited capacity of DNA sample identification and a lack of specialists. Information indicates that the only two centers capable of conducting DNA verification in the country can process less than 1,000 samples a year. But following the floods, more than 2,600 samples were brought in at once, which was far beyond the capacity they could contain.

Reports say as of September 11, 1,377 bodies had been recovered, but only 102 of them, representing seven percent, had been successfully identified and handed to their families for burial.

NowtrKit Founder Samar Panthi confirmed that so far, they have been able to identify about 90 persons from the more than 500 who have been registered with them. He explained that once the identification has been done, they pass over the information to the country’s police or other government agencies for verification before proceeding with the next steps.

The progress notwithstanding, the damage caused to many of the retrieved bodies makes both DNA and facial recognition very difficult, according to police spokesperson Abhi Narayan Kaphle. For DNA, officials have had to rely on teeth and bone samples which take longer time to process, while severely damaged faces have made facial recognition impossible in some cases. Researchers have developed a new system which can still be used to identify damaged faces.

An Australian Forensics scientist Prof. Richard Bassed told The Guardian early this month that many of the bodies retrieved may never be identified. He cited several challenges including insufficient resources and the difficult nature of the terrain.

The Nepal development comes as some have been advocating for biometrics to be increasingly deployed for disaster victim identification especially as an emergency response measure, as well as for the identification of dead unclaimed bodies in health facilities.

India used a facial recognition system to identify dead victims of a train derailment in 2023. Pakistan has, in collaboration with Chhipa, been able to identify thousands of dead bodies across the country.

Ukraine has used biometrics to identify the dead within the framework of its fighting with Russia, especially around border areas. In South Africa, health authorities can use a fingerprint system to identify abandoned or unclaimed bodies.

The problem with identifying dead bodies though biometrics

While the trend is growing, Nepal’s experience illustrates some of the concerns raised by recent research. That research, published in Big Data & Society, examined systems developed by Face Forensics, Clearview AI and Corsight AI, and found that the technology turns images of the dead into biometric data that can be stored, circulated and reused, including for training recognition algorithms.

The author cautioned that accuracy depends heavily on the quality and size of reference databases and the threshold used to determine potential matches. She also raised concerns about privacy, consent and the dignity of deceased people.  These are questions that require urgency when families are desperate for answers and authorities are under pressure to act quickly.

The tensions are somehow already visible in Nepal. NowtrKit says it passes identifications to police and government agencies for verification before any further steps are taken, a safeguard that acknowledges the stakes.

However, with more than 1,300 bodies recovered and only a fraction identified so far, Nepal’s turn to facial recognition reflects a broader reality that when DNA capacity falls short, biometrics may be the only option left even with lingering ethical questions.

Article Topics

biometric identification  |  biometric matching  |  biometrics  |  FaceTagr  |  facial recognition  |  Nepal