Trending: MacKenzie Bezos, Amazon CEO’s ex-wife, sells $400M in stock after pledge to give away billions
Henry Kroll demonstrates facial redaction of his own photo in real time, using open source software. Photos Bill Schrier
Henry Kroll demonstrates facial redaction of his own photo in real time, using open source software. Photos by Bill Schrier.

The Seattle Police Department (SPD) held its first-ever hackathon on Friday. The event was focused on a single problem: How to redact the video streams recorded by police officers from their dashcams and (soon) body-worn video cameras.

Seattle Mayor Ed Murray
Seattle Mayor Ed Murray

More than 80 people filled the room from 10 a.m. to 3 p.m. About one-third were technology professionals or part-timers like Henry Kroll, who makes a living as a salmon fisherman but focuses on video and other technology issues in his spare time.  The remainder were Seattle police and other public officials, a few members of the community, and a number of people from local companies such as Amazon Web Services and, plus a substantial media presence from local television stations and newspapers.

The event included cameo appearances from Mayor Ed Murray and Police Chief Kathleen O’Toole, both urging the participants to collaborate and solve the redaction issue. It was facilitated by SPD Sergeant Sean Whitcomb, who has led the department to nationwide leadership in the use of social media.

Mike Wagers, the SPD Chief Operating Officer, was very pleased by the results, saying they exceeded his wildest expectations, although admitting he had no specific expectations from the session. Wagers was the moving force behind the hackathon, which followed a public disclosure request for, basically, all the video recorded on the departments’ dashcams.  This request from an anonymous individual who tweeted under the handle @PoliceVideo and “came out” last week to identify himself as Timothy Clemans. Clemans was in the room and demonstrated his own index of disclosed video from SPD and other departments around the state.

Michael Mattmiller, CTO of the City of Seattle, helps to open the hackathon
Michael Mattmiller, CTO of the City of Seattle, helps to open the hackathon

During the hackathon, seven different teams or individuals either demonstrated specific tools and solutions, or outlined approaches they believed would work. Everyone agreed with Marcus Womack, General Manager of, who said that a purely automated solution was unlikely in the near future.  “In the end,” he said, “a human being will need to review sensitive video manually before it can be disclosed.”

The event started with Mary Perry discussing the parameters of redaction.   The Washington State public records act requires that almost all video filmed by any government agency – including police – be disclosed upon request.   The only real exception is for video which is part of an open case currently under investigation. However, various parts of the state code include other restrictions – the identity of minors cannot be disclosed.  Requests from victims or witnesses who may be at risk if their identities are disclosed also must be honored.  However in all such cases the video still must be released – it is just the faces or other potential identifying characteristics, which might include gender or even a person’s gait – which need to be blurred and redacted.

However, Seattle Police officials also admitted that about 90 percent of the video officers create probably needs no redaction at all. That’s because members of the public have no right to expect privacy in their interactions with police, unless they are juveniles or a witness or victim whose safety might be at risk if their identity is known.

Simon Winder demonstrates facial redaction of an Egyptian protest demonstration
Simon Winder demonstrates facial redaction of an Egyptian protest demonstration

Richard Li, Program Manager with Microsoft’s Azure Media Indexer Services, demonstrated that service. Video can be uploaded to the Azure Cloud and the associated audio will be recognized by Microsoft’s speech-to-text engine and displayed as a series of phrases next to the video.  Li stated that hundreds of video files – including the 350+ terabytes already archived by SPD – could be uploaded and audio-indexed.  Then the transcribed audio can be searched for keyword phrases. Much of the original audio is garbled or barely audible, and the service only recognizes English at this time, plus it has no capability of redaction either of the audio or people/objects in the video.

Simon Winder, a principal with Impressive Machines and formerly of Microsoft Research, demonstrated some tools for semi-automated redaction of faces in crowds. He showed a clip of demonstrations during the Egyptian protests of 2012-13 which showed replacement of faces with an icon.

A team of University of Washington students demonstrated facial blurring using the OpenCV library of tools.

Henry Kroll of the Nerd Show — he’s also the salmon fisherman – had one of the most compelling demonstrations.   Kroll used open source software such as the Blender and Pitivi to demonstrate blurring of faces and bodies in a video clip.  This blurring included movement of the individuals through the video.  Both tools show substantial promise.

Adam Monson, Lee Colleton and Phil Mocek advocate for an open source solution
Adam Monsen, Lee Colleton and Phil Mocek advocate for an open source solution

Womack, General Manager of, was next to talk is a local Seattle operation, although it is part of Taser International.   Although stores substantial amounts of video for its 3,000 customers – mostly police departments – all redaction is manual today.  Womack cautioned the audience to think “holistically”, for example, not asking officers to spend a lot of time entering metadata or uploading video – “they have better things to do.”  Although Taser and operate worldwide, its Washington customers are the most vocal, because the state has the most liberal public disclosure laws.

Clemans — the spark plug who ignited the whole event with his public disclosure requests — demonstrated his website which used the Azure Indexer to transcribe and catalog police video. Clemans also stated that most of the work done by police officers is saving lives through CPR or calmly working with drunken or angry citizens, yet the public rarely sees that on video.  He also urged a report where all video would be initially over-redacted – essentially the whole frame is blurred – and then the news media or activists could ask for less redacted versions of the video of interest.

The wrap-up team included Lee Colleton, Adam Monsen and Phil Mocek. Colleton is active in the Seattle Privacy Coalition. They urged SPD to consider open source software as a solution, and to partner with other agencies to fund such open source development. They also suggested the Seattle Police Foundation might sponsor such work, and partially fund it with a kickstarter campaign.

Next steps?

Some actions are happening quickly.  Seattle Police has already approved a policy for officer use of body-worn video cameras, and is starting a pilot today with 12 East Precinct Officers.

Where to take the hackathon results is less certain, but at least two potential paths are clear.  One is a partnership with a commercial company to upload and store video.  A pilot is in place with today.  Another path is the possibility of a coalition of police departments and foundations to develop semi-automated redaction of video to meet the conflicting requirements of public disclosure and protection of privacy.

Certainly Seattle’s police officers – as well as its citizens – are going to be seeing a lot more of themselves on video.

Like what you're reading? Subscribe to GeekWire's free newsletters to catch every headline


Job Listings on GeekWork

Find more jobs on GeekWork. Employers, post a job here.