Solutions for Law Enforcement
Responding to FOIA and Court Requests
More and more Law Enforcement Organizations are relying on captured videos to solve cases, arrest criminals, and document their work. The videos are sourced from a variety of places: body worn cameras, vehicle mounted cameras, store security cameras, commercial facility cameras, street cameras, citizen uploaded cell phone videos, internal archives, shared materials from other law enforcement agencies, etc. Officers use these materials in the course of their investigative work leading to cases that can be prosecuted to improve the safety of the community. When a court requests evidence, LEOs are typically given a short amount of time to do so (10 days generally). Often image/video evidence contains material that is not pertinent to the case and that should not be included in order to protect privacy of the bystanders, or other sensitive information (for example, the face of an informer or perhaps a witness also visible in the shot). The amount of video that needs to be provided in redacted form has been growing in recent years.
Some software solutions that deal with video redaction are currently available today, however the results have been of mixed success. Redaction errors typically involve missed faces (not blurred and therefore fully visible), which therefore have to be handled manually by using a video editing program. This is very time consuming and not practical in most cases. Furthermore, available solutions are often bundled by body-worn camera suppliers and are limited to processing just those types of files, and not files from other sources. As a result, police departments often rely on outside service providers for redaction needs. The videos are sent to the outside company where it is manually processed. The output quality is generally very good, but the costs are extremely prohibitive. Also, the more complex the redaction specifications are, the higher the costs are, as it takes more time for an editor to complete the work. Given the limited time available to respond to a court request, scheduling can pose challenges in many situations.
redactX transforms the overall experience dramatically:
redactX is the most accurate in the industry. In most cases, the automatically generated redaction output is typically “good to go”. In situations where something is missed, a user can fix the problem with just a few mouse clicks. The software learns from that input and generates an updated and fully correct output file.
The software covers redaction requirements more thoroughly than any solution on the market. Over 100 image/video formats can be processed (the source of the file does not matter as long as the storage device (a local disk, or a mapped drive) is accessible to the software. Data in S3 buckets and in Hadoop clusters can also be processed. The level of image analysis carried out by redactX is also greater than what has previously been available. For instance, other solutions might work only with frontal faces, redactX works with faces from any angle and rotation. In addition, it also processes heads from any angle, including the top, sides, and back. Distinguishing marks (a ponytail, a bold spot, etc.) are de-identified more thoroughly than what could be done with a face-only blur. Within reasonable bounds, the presence of glasses, sunglasses, beards, partially occluded faces, small faces, not well lit faces, etc., are all conditions that the software can accommodate readily.
It is very flexible, and redacts not only faces/heads but also text strings (in over 90 languages) and objects (“person”, “car”, “gun”, etc.). The level of flexibility is further enhanced by letting users choose whether all faces should be blurred, or only some and, if so, which ones. In the case of text, users can ask the software to redact all text strings, or only some text strings. The software also allows for partial string matching so that any word containing the indicated sequence of characters will be blurred. redactX can handle Latin-character based languages and also ideogram based languages and other international character sets. License plate numbers are treated as any other text string. Users can set up, save and recall/reuse, templates that specify what is to be redacted and how the output should look (format, video bit rate, choices for pixelizations, blurs, etc).
Law enforcement organizations can install the software on-premises or run it as a machine image on AWS including the Fed-Ramp certified government version. Either way, they retain full control of the media that is processed and who can access it and how. Whether on-prem or in the cloud, redactX is a fully encapsulated application that runs in both Windows and Linux environments. The software is also highly scalable and can be used in a distributed compute environment with a redact-X “manager” server controlling multiple redactX “worker” servers. GPU acceleration is desirable, but the software can also run in a laptop environment if speed is not a criteria for the user. The user interface is through a standard HTML-5 compliant web browser. therefore multiple users can login and work with the system concurrently. The customer’s “System Administrator” controls and assigns login credentials.
redactX is the first product on the market with image/video content database support. redactX users can set up reference databases of faces, such as for example mugshots of criminals, and use that reference as a filter. In that example redactX can be setup by the user to automatically redact anyone who does not match a face in the mugshot set. redactX has a REST-based API that enables additional functions, so the application is customizable to fit a range of customer criteria and modes of operation.
There are a large number of use cases that redactX can support. Please contact us to get more information.