Multi-object detection – now in Griffeye Brain

Multi-object detection – now in Griffeye Brain
August 28, 2019 Sara Ekberg

Multi-object detection – now in Griffeye Brain

Ever since its launch last summer, our AI technology, Griffeye Brainhas been a success in changing the way investigators approach CSA investigations. After the first release it has been continuously developed in several steps, and now it’s time to take the next big step. Today, we release the full version of Griffeye Brain Objects.  

The very first release of Griffeye Brain was focused on the CSA Classifier, which scans through previously unseen footage and suggests images that it believes depict child sexual abuse content. Many of our users have been testing it in various investigations – and some have even implemented it as part of their daily work with really good results.

Now, we have also developed and upgraded the beta version of the object classifier to a full version. Griffeye Brain Objects is not only faster and more accurate, but it also has the capacity to identify multiple objects in one single image.

Adding another dimension of intelligence

One of the biggest challenges that law enforcement faces is uncovering details that could be potential evidence. Investigators often need to work in the dark, without really knowing where to start and what details and visual attributes to look for in masses of previously unseen data. This is one of the main reasons why we developed Griffeye Brain Objects. By using AI technology that is trained to detect different objects and details in images, users will get an indication of where to start investigating and what data is relevant to the case.

Highlight multiple objects and find relations

The Griffeye Brain Objects automatically detects objects and details at import, and then bookmarks and sorts the images into a hierarchy of classes with different main concepts and sub-concepts. Let’s say that the algorithm detects a car or a bus. Then the image will also be labeled into the main concept ‘Vehicles’. By sorting images this way, users can now filter and navigate their case in a way they couldn’t before. And by locating and labeling different objects and details in images, it will also be easier to identify relations and link correlating images based on visually similar or matching objects.

Enable more clues to be found

By filtering out relevant information with the help of AI technology, users will speed up their investigations and free up time to prioritize investigative work. With an object classifier added to the Griffeye Brain, users will also be able to find and connect more critical clues and identify relationships that otherwise could have gone missing.

“Allowing the computer to pre-classify objects allows investigators to locate relevant content exponentially faster. Tasks that used to take several hours or even days can now be done in seconds. Adding to that, it also improves workflows and speeds up victim identification efforts. So, you could say that everybody wins.”  – Eric Oldenburg, Law Enforcement Liaison, Griffeye


READ MORE: Release of Analyze 19.2 – Boosting the AI power

More Griffeye news

  • Release of Analyze 19.5

    Analyze 19.5 is now available! This release is mainly about maintenance and usability improvements…

  • Join us and Bellingcat at ICDDF 2020

    March 16th-20th, we will be at the ICDDF Conference in London, UK. On Monday 16th we will be hosting a workshop…

  • How AI and Facial Recognition Are Transforming Police Case Management

    Johann Hofmann, CEO of Griffeye, discusses why AI and facial recognition are…

  • Release of Analyze 19.4

    In the 19.4 release, we have mainly focused on improvements to the Griffeye Intelligence Database (GID). But we have also…

  • Webinar: How to leverage GPS data in your digital media investigations

    January 16, 2020 @ 15.00 CEST (9 am EST) and 4 pm PST (7 pm EST)

  • Identifying victims in millions of images and thousand hours of video

    While data contains a wealth of valuable information that previously would have been impossible to find…