New Challenges With Body Camera Footage

Chris Cruz gave a very insightful lecture into a topic many students, including myself, may be a bit underinformed about: technology use in government. Before this lecture, I assumed the technology used within the government would be extremely outdated and basic, however, Chris gave evidence that points to the opposite. For example, the centralized cloud infrastructure that is used within all departments of the California government. Chris mentioned that the centralized cloud is important for accessibility as well as security, which are definitely two advantages of utilizing the cloud. Another important topic was making high value data sets available to utilize tech to serve Californians. A question was asked in lecture about camera usage in California, and Chris mentioned there is a push to put a body camera on every police and California Highway Patrol officer, however, this poses a great technological challenge with introducing all of that new data. The camera footage should be included as high value data sets, so not only does this footage need to be stored efficiently, it needs to be readily available as well. In a state as big as California, there will definitely be an incredible amount of data gathered per day from body cameras.

Some cities, including some within California, have already began utilizing body cameras regularly in police forces. An interesting example I found described the Buffalo police force, and usage of body cameras outside of the Bills stadium since there is a higher chance for crime around the stadium due to the football games [https://gcn.com/articles/2017/05/11/body-camera-storage.aspx]. As known in current data centers, data input and output expectations are not static; the demands can vary per hour, or even per minute depending on the use case for the data. For example, the usage vs time chart of Google’s search data centers per location would peak during the day, but remain relatively low at nights due to the nature of the workload and the user inputs. However, after a major event, the usage can spike heavily and even go beyond what the data center is capable of. On this note, the article I quoted mentions how fall Sundays can lead to an incredible spike in data due to the Buffalo Bills football games. The Buffalo police force not only needed to store body camera footage, but also  “needed to store both data from mounted surveillance cameras and operational data from town employees” and to do so they turned to software defined storage (SDS). (gcn.com) Utilizing SDS allows the police force to use industry standard hardware by decoupling the physical storage hardware from the storage tasks, and performing those tasks in software. The article also mentions that SDS and similar solutions are continually developing from open source communities, which allows users to have the most cutting edge solutions by fostering innovation and collaboration.

Once data is stored efficiently, it is also necessary to be able to utilize that data. Again, with a state as large as California, there will be a massive amount of new data introduced per day, and there are great learning opportunities that arise with this new data set available. However, using actual man-hours to view and parse this footage would be unfeasible as data would be input at a much higher rate than what could be viewed. This is a great application for machine learning and artificial intelligence, which are areas where Stanford excels. Very recently, Stanford researchers performed a study on body camera footage and found “police interactions with black community members are more fraught than their interactions with white community members.” [http://news.stanford.edu/2017/06/05/cops-speak-less-respectfully-black-community-members/]. This study was performed by forming an interdisciplinary team from computer science, linguistics, and psychology departments, and analyzing transcripts from Oakland Police Departments body camera footage. The researchers were able to develop a machine learning model that could detect aspects of speech such as respect, apology, concern. The article also concludes by expressing how much potential there is from analyzing body camera footage, and the implications it has on bettering the community.

It will be very interesting to see not only what kinds of technology is utilized to efficiently store and process this massive new data set, but also the legislative side of storing all of this data. Some data will need to be stored temporarily, and some will need to be permanently; some data will be available for all to see, and some will need to be protected as confidential. This intersection of legislation and technology is exactly where experienced employees such as Chris Cruz will be necessary, and it is exciting to think of the possibilities in the future if all challenges are solved appropriately.

Users who have LIKED this post:

• 
•