Video redaction software enables enhanced transparency for Ohio police department
When the Chief of Police in Oregon, Ohio came to Sergeant Jason Druckenmiller about how the force could increase transparency by making video evidence public, they immediately ran into a problem. Various laws and other considerations make it necessary to redact various kinds of personal information from any publicly-released evidence, and the department did not have any obvious way to do so.
There are many ways of approaching the issue used by police forces around America, and law enforcement agencies vary greatly in how much material they release. Ohio introduced laws governing the balance between privacy and transparency for body camera footage last year.
Without any obvious way to carry out redaction on large volumes of video, and lacking the resources to simply assign a team to combing through every video taken or received by the department, Druckenmiller told the Chief that the force would be limited in what it could release.
“Every opinion that I’ve gotten so far said if we don’t have the capability to do it then we can just hold that record,” Sgt. Druckenmiller explains to Biometric Update in an interview, “and the Chief said ‘No. We’re not going to hold an entire video just because four seconds of it need to be redacted.’”
The Chief’s first suggestion was to hold a piece of paper over a certain area of a screen while filming it with cell phone.
Sgt. Druckenmiller was aware of software that could potentially help, however, and looked into the solution provided by the force’s digital evidence technology provider, Coban (now part of Safe Fleet), as well as others on the market. He ran a test on a ten-minute video the department had of a night-time traffic stop against the different options.
What he found is that some of them use the same engine on the back-end. The Veritone Redact solution, however, uses a different one, which processes video much more quickly while more accurately identifying the objects and areas that need to be redacted, and allowing changes to be made more easily.
Veritone also offers face biometrics, but the Oregon Police Department is part of a regional network which shares booking photos from different departments and a facial recognition capability.
The advantages of its video redaction software were enough to make it compelling for Sgt. Druckenmiller and the Oregon P.D.
“For example, if an officer is holding a driver’s license in front of their body cam, that personal information needs to be redacted,” Sgt. Druckenmiller explains. “The personal information isn’t going to get picked up by the engine that would determine that it needs to come out, so we need to manually, basically draw a box around that and track it through the picture. Well, a lot of officers tend to talk with their hands, so as that driver’s license is coming in and out of the frame and moving around in the frame, I’m able to just draw the box and hold my mouse over it, and I can move my mouse with the movement of the driver’s license in the picture, and I don’t have to go frame by frame and draw a new box on every frame as that object moves around.”
The solution also has to run on government equipment, so even though as a cloud solution the department is somewhat at the mercy of the network’s internet speeds for video ingestion, it avoids the need to invest in a high-powered PC with a specialized video card.
Sgt. Druckenmiller gets around the lengthy upload time which is the drawback of Veritone Redact being a cloud-based system by typically starting process when he is leaving work for the day. That way it is completed the next morning, uploaded and processed and ready for any adjustments he needs to make.
Manual adjustments are inevitable, according to Sgt. Druckenmiller, because of the nature of the task. While he considers himself a proponent of automation, it can only go so far.
The reason for this, he explains, is that it is not always possible to tell what should be redacted without fully understanding the context. One item on the list of what should be redacted is images of a deceased person, or the death of a person, unless that “death is caused by peace officer.”
“How are you going to automate that?” he asks. The engine picks up the person’s face, but not whether the person is alive or dead, let alone the circumstances of their death.
Automating more identifications with forensic facial recognition could help the Oregon P.D. further protect people’s privacy, Sgt. Druckenmiller says, as the force in some situations resorts to posting images of suspects or other people of interest online and asking the community to help identify them. In some of these cases, he notes, the people have not done anything wrong. Even if the individual has committed a crime, if police were able to find out with a facial recognition match, rather than by appealing to the public, negative impact on that person’s privacy can be reduced.
Ultimately, Sgt. Druckenmiller argues, transparency is the key.
“The more people understand why we’re doing what we’re doing as law enforcement, a lot of times the better the interaction is,” he says.
Oregon, Ohio’s Chief of Police wanted transparency, and the law and courts place certain things out of bounds. The video redaction software is what allows both needs to be met.
“Transparency in law enforcement goes a long way towards building that trust with the community, and by us having a redaction capability, and one that doesn’t take up a lot of time – I don’t have to have an entire staff of people that does this – I can make sure that these requests get fulfilled and a lot of times get fulfilled that day or the next day, and we can get that information out right away,” says Sgt. Druckenmiller.
The speed and efficiency provided by the automation, after all, is necessary to providing that transparency in time for it be meaningful.
“It doesn’t do me any good to be transparent six months later,” Sgt. Druckenmiller points out.