AI tech company Veritone has unveiled a new AI-powered tracking solution that leverages confidence-based similarity detection to track people and objects through video recognition software. Veritone Tracker provides the unbiased insights needed to assess the visual description of a person of interest in a scene, what they are doing and the context of others in the scene, without performing facial recognition or other biometric identification that would reveal a person’s identity.
Veritone Tracker was created using the company’s aiWARE and AI domain knowledge of evidence lifecycle management, as well as technology from its August acquisition of London-based Vision Semantics Ltd. (VSL), a leading computer vision technology company focused on AI-powered video analytics and surveillance software solutions. The combined technology, says the company, will enhance its existing aiWARE-based applications in the public safety and commercial enterprise markets.
VSL has been internationally recognized for its Person Re-identification (RE-ID) solution, a privacy-preserving computer vision technology for public safety and security applications. Person RE-ID is the mechanism to find a person at different locations over different times in a vast quantity of video data collected from distributed cameras. It’s unique because it doesn’t use private data, facial imagery features or other commonly employed person-specific biometrics.
“Integrating VSL technology into aiWARE enables us to expand our capabilities and provide new offerings for existing and future customers,” says Ryan Steelberg, Veritone president and co-founder. “Veritone Tracker is the first example of this, using VSL’s renowned RE-ID technology, which can be applied across a variety of use cases. With RE-ID, we can enhance our current applications to help locate individuals without compromising an individual’s privacy.”
Veritone Tracker enables corporations, governments, law enforcement agencies and justice organizations to quickly build an understanding of an individual’s activities and associations by assembling a collection of moments in video files to create a timeline. This type of “digital forensics” helps build a narrative around an event or an individual that is captured on video. Veritone Tracker can also help the private sector, including arena/stadium event teams, private security, daycare staff, elderly care centers and property management groups assess on-site video footage to find missing people or investigate specific situations that occurred on their premises.
“This new technology enables organizations to focus on the work that matters most, expediting resolutions and keeping neighborhoods safe, without compromising private data,” says Jon Gacek, general manager, aiWARE Enterprise, Veritone. “It also has broader use cases outside of law enforcement that can help a variety of organizations across many industries conduct their own investigations into video-captured events and protect staff, patients and employees more effectively and efficiently.”
In any type of investigative situation, whether conducted by a justice or public safety agency or another party outside the jurisdiction of the judicial system, video surveillance and evidence review is often performed manually. With the wide use of video enabled applications and many video enabled platforms, it can take countless hours to understand where and when people appeared and the context of a scene. Veritone Tracker streamlines this process, preserving valuable time and resources to achieve the best outcome and keeping citizens and the community safe.
Veritone Tracker’s AI engine identifies people as “human-like objects” (HLOs) in frames across videos, and stores those as detections. Tracker can return potential matches from those HLOs identified by the AI engine that best match a person of interest (POI). Tracker can also apply that process across a collection of videos and return all occurrences of a POI for review by a user.
Veritone Tracker can make it easier for investigators to identify a person of interest in a media file so that the features of that individual can be surfaced across other ingested video files from multiple sources and camera angles. They can use this to help find a missing person faster and build a timeline of events across traffic cameras, security footage, and other video sources that would otherwise require manual review to piece together.
Artificial intelligence helps make video evidence review more effective, efficient and economically competitive; however, technologies such as facial recognition have traditionally brought privacy concerns to the forefront. Veritone Tracker, says the company, addresses many of these privacy concerns while enabling agencies to keep communities safe while managing the burden of reviewing hours of video footage.