Contact form
Your request has been sent.
Please enter your name
Please enter your email
Please enter a correct email address
Please enter your company name
Please enter your message

DASA case study - Accelerating Artificial Intelligence (AI) on the Front Line

← All resources
Use cases
DASA case study - Accelerating Artificial Intelligence (AI) on the Front Line

04/01/2021 by Preligens

Preligens harnesses deep algorithms to aid timely military decision making

Military analysts must wade through vast amounts of images and data to identify patterns in target and threat areas. This takes time, and time is a precious commodity in hostile situations. Preligens was funded through DASA to mature its AI technology to aid operational intelligence gathering.

It is the labelling of the images and data sets that takes up much of a skilled analyst’s time, but it is a vital process if the military is to build, for example, a pattern of adversaries’ behaviour. Preligens is helping to automate this process by advancing, more quickly, the deep learning algorithms through its state of the art Artificial Intelligence and machine learning.

Throughout the project Preligens worked closely with military operators, decision makers and technical experts at the Defence Science and Technology Laboratory (Dstl). Such unique inclusion into the military world, allowed Preligens access to ‘real life’ scenarios, data and satellite images. This created a stream of high quality data, ensuring accurate analysis. Dstl Technical Advisor said “It has been a complimentary active learning project where the military and ourselves have learnt from the research alongside Preligens.”

This innovative research has sought ways to create a mechanism to find, from amongst all the new images available, those that will make algorithm performance progress the most.

The team successfully demonstrated that algorithm performance is improved rapidly through labelling only certain data and that considerably less human effort is needed to label it. They also proved the power of active learning for improving the performances on poor-quality imagery, thereby paving the way for a maximally automated improvement cycle for the algorithms. On the back of the research, Preligens will use key data to inform their future software maps and proof of concept.

“The DASA funded ‘Active Learning’ project has accelerated our path to commercialise this technology. The access to ‘real life’ scenarios has enabled us to refine our technology and our understanding of the military user, so that we can build an innovative solution that is applicable and immediately useable in this sector. The knock-on effect to other commercial markets is also felt - as we can use our learnings to make refinements to other products." Mathieu Goebel, VP Sales, Preligens

Related articles
A concrete AI application: Multi sources data analysis to localize planes
A concrete AI application: Multi sources data analysis to localize planes
A concrete AI application: Multi sources data analysis to localize planes
How hard is it for an AI to detect ships on satellite images?
How hard is it for an AI to detect ships on satellite images?
How hard is it for an AI to detect ships on satellite images?
Boosting object detection performance through ensembling on satellite imagery
Boosting object detection performance through ensembling on satellite imagery
Boosting object detection performance through ensembling on satellite imagery