Security + Compaile
High recognition rates > 93 %
Improved management of resources
“In a world where information is becoming increasingly important, faster, and more diverse, it’s essential to be able to keep track of it all. The solution COMPAILE created doesn’t just focus perfectly on one aspect of the challenge; it covers the entire image/video analysis process and never loses sight of what the user actually needs.” - Anonymous, German government agency
The work of our security services isn’t always exciting. Sometimes, it involves sitting in a spartan room for days, staring at a computer screen. And the workers aren’t writing reports or participating in video conferences; they are clicking through thousands of images or hours of video footage as part of the grueling task of video and image analysis.
One consequence of digitalization is that growing numbers of files are being seized as evidence. Sometimes these files are images from a confiscated smartphone, sometimes video recordings from a surveillance camera. When exactly did the suspect’s Audi pull into the gas station? Was someone running around the city with a handgun on the day of the attack?
Most security services do not yet have access to automatic object recognition software, which can analyze an image and distinguish a weapon from an umbrella, or a swastika from another symbol, for instance. And only a few agencies have the personnel numbers needed to handle the time-consuming process of analyzing images and videos. There’s also another problem: “Identifying relevant information in large volumes of data – particularly in images or videos – isn’t just time-consuming; often, it is also practically impossible for the human eye,” says Ayse Gashani. This is particularly true if the data has to be assessed quickly, such as in situations involving kidnappings or fugitives from justice.
The solution developed for this case saves investigators many hours of screen time and allows them to focus on the most worthwhile approaches to fighting crime. The solution uses artificial intelligence to scan large volumes of image data and marks relevant sections for the analysts to assess. In the process, the software utilizes neural networks and machine learning – so it recognizes patterns independently. Consequently, analysts no longer need to view the entire data; they can simply assess the sequences identified by the system.
A number of security services have already used the software successfully.
Our solution automatically identified and marked objects such as vehicles, weapons, and symbols in the images and videos, making it easier for investigators to assess the prioritized images quickly.
The results are enriched with metadata belonging to each file, which is another helpful factor. The solution automatically reads out this metadata, as well. The data can also be exported or integrated into an existing system environment via an API for further processing. Additionally, analysts have access to a web-based analyst interface, making the process of analyzing images and videos easier, faster, and more precise.
THE SUCCESS STORIES
Over 3,000 objects can be recognized automatically, allowing organizations to use resources more efficiently, as employees can concentrate on relevant data.
more success stories
Component recognition and quality assurance based on a content-driven similarity comparison at TRUMPF Werkzeugmaschinen GmbH & Co. KG
Entity recognition and media management based on a neural network in a web-based user interface at German government security agencies (domestic and international security)
Open, distributed, real-time-capable, and secure operating system for data-driven, AI-based applications in production.