Seattle Police Hackathon Tackles Video Transparency

The Seattle Police Department is leading the nation in police transparency -- with a little help from the public.

by / December 22, 2014
YouTube/Police Video Requests/Tukwila Police Department
YouTube/Police Video Requests/Tukwila Police Department

This collection contains more than 1.6 million movies – 314,000 hours of footage spanning 360 terabytes of disk space ... and it's growing.

It’s not Netflix – it’s the Seattle Police Department’s (SPD) video archive, generated by dashboard cameras mounted on the city’s patrol vehicles. The department burns about 7,000 DVDs each month to fulfill public disclosure requests for video, and it takes a dedicated team of five employees who do nothing but burn DVDs to meet the state’s aggressive transparency laws. The process of reviewing, redacting and releasing the video is expensive and time-consuming, and that’s why the department held a hackathon on Dec. 19 -- to search for a better way.

More than 80 attended the SPD event, including public officials, police officers, students, technologists showcasing open source solutions, and representatives from companies like Evidence.com, Amazon and Microsoft. 

The objective for SPD is to automatically release all police video online, and SPD Chief Operating Officer Michael Wagers said the department's policy – still in development – will be a model for the nation, but it needs the support of new technologies to work. The formation of new laws, policies and technologies is particularly important given the growing use of body-mounted cameras by police. SPD launched its own body camera pilot using 12 officers on Dec. 19, which will end in the spring.

This gathering of local technologists didn’t conclude the search for new technology in releasing police video, but it gave the department some new ideas and it's now deciding what to do next, said SPD Spokesperson Drew Fowler.

“The project of our redacting video is very fine-tuned and it requires a lot of attention,” Fowler said. “So at this point we are definitely not able to [abandon] the human element, but we’re looking at some other ideas that might get a larger amount of video out sooner, but it would be of lower quality. We’re playing with a lot of ideas that might ultimately be incorrect, too, but hopefully we’ll be able to streamline the process and answer more questions, because as it lies, it’s impossible to redact the amount of video that is being requested of us in a reasonable timeframe.”

Though SPD called the event a hackathon, there wasn’t any coding going on -- it was really more of a series of presentations by companies and software developers, said Timothy Clemans, one of the software developers who presented at the event. Clemans is helping the SPD develop its public records disclosure and publishing policies. Clemans is the formerly-anonymous character known as the Police Video Requester, who precipitated much of the discussion around public records policy in the state.

At the event, Clemans presented a project he had developed using a new audio transcription tool developed by Microsoft, called Azure Media Indexer. In one demonstration, Clemans shows that it’s possible to remove a video’s audio, turn the video to black and white, blur each frame of video, and accompany the video with a time stamped audio transcript. Clemans calls this approach a middle ground proposal because it’s still possible to understand most of what’s happening in a given video, but it affords those in the video a degree of privacy and solves a logistical challenge by reducing file size by 75 percent.

In another video – a nearly six-hour interrogation video involving two accused of a recent school shooting – the tools demonstrate how video content can be reduced to a more text-based accessible format, inaccurate though it may sometimes be. Releasing data in this format is not only more accessible to the public, but integration with redaction tools can allow officials to easily redact segments of audio by clicking on the individual words they want to redact before releasing the content.

SPD is now pursuing a contract with a cloud provider to begin putting its videos online, Clemans said. The videos are to be compressed to 8 percent of their original size, which would reduce the department’s collection to a still-large but more manageable 29 terabytes. The question that remains is how SPD and other departments will manage the redaction process. In SPD’s case, about 90 to 95 percent of the video could be released without any need to remove information to protect anyone’s privacy. The problem, however, is that the videos aren’t tagged, so no one knows which videos are the ones that need redaction.

SPD’s new police body camera program will circumvent the redaction problem by requiring officers to tag any videos requiring redaction from the start, and eventually those videos will be automatically uploaded to the Internet. In the meantime, however, SPD and many other police departments need ways to automate the redaction and release processes.

Former Seattle Chief Technology Officer Bill Schrier reported that no one in attendance anticipated a completely automated solution in the near future. Clemans agreed that a more perfect solution than the kind he presented is many years, or at least months, off, which is why he is encouraging the city to pursue imperfect solutions using today’s technologies so that large quantities can be released as soon as possible.

While there are not yet formal plans for the SPD to pursue any particular project, the department is now seeking funding for a project wherein a group of open-source developers can create a suite of redaction tools that could be used by agencies worldwide, Clemans said.

Clemans himself will meet with Evidence.com later this month, he said, to collaborate on the development of some video over-redaction tools.

Colin Wood former staff writer

Colin wrote for Government Technology from 2010 through most of 2016.

NEW ON THE PODCAST

Tech Startups Can Help Revitalize Both Ends of Your Ballot