Taser acquires a computer vision startup to bring AI to police body cameras

With a pair on acquisitions on Thursday, Taser hopes to soon use artificial intelligence to solve a problem increasingly faced by police departments: sorting through the growing torrents of video recorded by body cameras to find significant events, objects, and people.

As part of the deals, Dextro, a New York City-based computer vision startup, along with a team of computer vision researchers from Fossil Group, based in Richardson, Texas, will form a new department within the company’s camera-focused Axon division called Axon AI. Taser did not disclose financial terms of the transactions, which will bring nearly 20 researchers and engineers to Taser’s Seattle-based, 110-person Axon unit, along with technology to help police “quickly isolate and analyze the most important seconds of footage from massive amounts of video data.”

Above: a slide from a 2015 Dextro investor presentation

Scottsdale, Arizona-based Taser already commands the country’s market for police electroshock weapons and leads a growing body-camera industry that has sprung up amid concerns about police behavior and accountability. (It has also attracted criticism from competitors for an aggressive approach to securing police contracts.) The acquisitions are intended to bolster CEO Rick Smith’s vision of “smart” body cameras that can recognize objects and people and eventually integrate with automated police records and evidence management systems.

“We can already pick out when guns are present or when there’s a protest going on,” David Luan, the company’s founder and CEO, told Fast Company last July about another project, helping moderators scan online video for offensive content. With the acquisition, Dextro will now be able to train its algorithms on the 5.2 petabytes of data Taser’s customers have stored in the cloud, which Luan said in a statement was “the largest repository of law enforcement footage and annotations in existence.”