The Challenges of Tracking Persons of Interest Across Video Files (Part 2)

How Veritone Tracker Accelerates Investigations Without Violating PII


  • Analyzing video evidence has historically been a laborious, manual process for investigative teams. 
  • AI technology that uses facial recognition has the capability to automate some video evidence reviews, but it’s not great with crowds and raises concerns about privacy.
  • New AI technology that doesn’t use facial recognition can now rapidly accelerate video analysis when searching for persons of interest across different files. 


The first part of this series discussed how video evidence has become pervasive across most investigations. This largely has been attributed to the increased adoption of body-worn cameras across law enforcement agencies and the proliferation of smartphones and devices. As such, video evidence review teams manually review hundreds of hours of footage when it is against them if perpetrators are still loose or if they need to understand a crime to develop a more cohesive case.

However, using AI for investigations raised questions about privacy laws and maintaining personally identifiable information. Specifically, how could one use facial recognition technology without violating these areas, thereby damaging community trust? With advances in new AI technologies, facial recognition is no longer needed to track persons of interest across video files. But how does it do that? 

A More Ethical Approach to Tracking Persons of Interest with AI 

New technology uses a new method called confidence-based similarity detection. It can detect and track people and objects in a video without performing facial recognition or any other type of biometric identification that would reveal a person’s identity. Through this technology, investigations can move faster than ever before. 

Why does time matter? In missing person cases, investigators unanimously agree that the first 48 hours are critical. Evidence is still fresh, and the chance they’ll be found is high. Three-quarters of missing persons are found within 24 hours. So using AI to expedite video evidence review  greatly increases the chance of finding a missing person. That same speed can also be applied to tracking down and apprehending criminals. 

But teams quickly become overloaded with the vast majority of investigations having some form of video evidence and no way to automate its processing. And even if they wanted to use facial recognition, if there are large crowds of people on camera, making faces hard to ascertain, that quickly becomes unviable. That’s where human-like object (HLO) detection comes into play. 

Identifying People as HLOs with AI

Human-like object detection is when an AI engine identifies people based on certain known features it’s trained to look for. Such as clothing, piercings, shoes, etc. The AI can then surface moments where it thinks these HLOs appear, making it more time efficient when reviewing hours worth of video footage. At the same time, it overcomes the shortfalls of facial recognition and reduces concerns about privacy and exposing PII. 

Serving law enforcement agencies already with our portfolio of offerings, Veritone recognized the need to accelerate investigations. Enter Veritone Tracker, a digital forensics tool for law enforcement, public safety agencies, and building or event security teams to analyze videos from separate sources to track persons of interest. 

Using the technology described above, Tracker removes the need for a manual review of video evidence to build timelines and comprehensive stories for investigations. Whether you need to better understand an altercation at an event, find a missing person, or track and build cases against criminals, Tracker can accelerate any investigation. Used in conjunction with our other AI capabilities, organizations in the public and private sectors can turn the clock back, gaining invaluable time that can directly impact the outcome inside and outside the courtroom.

Read Chapter 1

Learn More about Veritone Tracker



Further Reading: 

Orange Police Department on Veritone Contact

Oregon Police Department Selects Veritone Redact 

Redaction in the Digital Age: Here’s What You Need to Know