Suspect Technologies provides an AI platform to help law enforcement agencies accelerate and enhance their ability to protect, investigate, alert and report.
As an ADF Partner, Suspect Technologies brings an additional power of analysis to ADF's suite of triage and digital forensic software to support face recognition for age detection which is used to help speed child exploitation investigations.
Brittany: How did your time as Physics Researcher and Research Fellow for Neuroscience and Research Computing shape your thinking about facial recognition? Or did this experience allow you to go into researching AI at Princeton?
Jacob: It gave me a good background in Artificial Intelligence than anything. We use very similar tools in neuroscience that we use in AI so it’s a lot of pattern analysis and making sense of large data sets so it’s very similar.
Brittany: What was your main motivation for entering the computer vision/facial recognition field?
Jacob: Well I’ve been working on imaging my whole career. I’m a physics person from academia and we work on biological imaging and such. We answer large-scale questions like how we go from egg to embryo. It’s all imaging and that’s been my whole career.
Brittany: Why did you start Suspect Tech?
Jacob: We started Suspect Tech with a specific use case in mind which was the redaction technology. We got the quickest traction in the security space with the redaction software. It started as a few demos and then gained some traction. All in one month we closed our funding and scored a contract with a large camera company and formed the company (Suspect Tech) all in the same month.
Brittany: Can you talk about Suspect Technologies' role in working with other organizations/partners?
Jacob: We help different organizations with their visual analytics needs and a lot of it is facial analytics. Some things are related to that. Recently you know people had a lot of issues ingesting different file formats for example so we send the platform in a few ways like that you know. So essentially our company has a redaction technology, investigative analytics, which we owe some of that to ADF, and enterprise/security around access control.
Brittany: What were your biggest challenges with creating the AI software? (“developing a facial redaction product for companies that produce body cameras”)
Jacob: Yeah you know we’ve been through a lot of revisions. Redaction was definitely one of the toughest. The ability to tag and track very small faces. We spent 18 months on an update which was The small face update which allowed us to do smaller pixel faces which is important for body cameras when a person is walking up from a long distance and their face goes from very small to very large. That was one key thing definitely. Even with good technology, it takes time to redact. With ADF, even if we get 80 to 90% of the child exploitation images/frames correct but not everyone is satisfied with that and it takes time to do that.
Brittany: How has covid-19 affected your facial recognition technology? How do facial masks or other occlusions (hats, sunglasses, partial profiles) affect facial recognition software? Does it work just as well?
Jacob: We have definitely optimized for those occlusions: hats, glasses, and such. It definitely required some updates. On smaller datasets, the results are pretty good. On larger datasets, it can be a bit more challenging. For masks for example the technology can tell that it is the same person with or without a mask. We have optimized for that.
Brittany: How much has facial recognition accuracy improved in recent years?
Jacob: It definitely has improved a lot. I don’t remember the numbers from the top of my head. The results are good enough to be used for a lot more use cases. Facial recognition is being used in airports and cellphones so it’s definitely come a long way. A lot of applications (thousands of people) you see really good results and that’s a lot of faces to track and link.
Brittany: How easily do you integrate with video management systems?
Jacob: It depends on the video management system. We have more evidence management partners than video management partners in the law enforcement space. Usually, we can integrate the technology as an iframe embedded into a portal. We can do integration in 2 days or less. Some video management systems can be a bit more difficult to integrate with than others. It’s pretty plug-and-play.
Brittany: Do you know if Suspect Tech reps or yourself are attending any more forensic conferences this year or plans for next year?
Jacob: We will attend a number of law enforcement conferences, but not as many forensic conferences. I was just at a conference in Rhode Island that involved law enforcement. I also attended the Techno Security Conference in Myrtle Beach.
Brittany: Are there any exciting plans for Suspect Tech in the future?
Jacob: We are working on a new use case which is domestic violence. There are more partners integrated with Suspect Tech. We are looking for any opportunities where it makes sense to work on.
Brittany: What do you like to do in your spare time?
Jacob: I spend a lot of time on my start-ups. I also golf, volunteer at an animal shelter, and read occasionally.