Visimo, alongside research institute partner Colorado State University, has won a US Air Force (USAF) Phase II Small Business Technology Transfer (STTR) award.
The award will see Visimo develop a solution that will help tackle training data problem for the Air Force Research Laboratory, Information Directorate (AFRL/RI).
Visimo’s technology will support the AFRL Lab in the development of several technologies.
The company will design a generator that can build synthetic annotated image datasets within minutes.
Visimo chief data scientist Dino Mintas said: “Specialised, large-scale datasets are required to train machine learning models.
“The advancement of ML and what these models can accomplish are often slowed due to a lack of appropriate training data.”
Visimo is designing a Conditional Generative Adversarial Network (CGAN) that will learn to use measured volumes of landscape images to generate ‘unlimited amount of original backgrounds’.
The Phase II STTR will focus on expanding the number of specialised output parameters that can be created by the CGAN, including security features and biome changes.
This award follows the success of Phase I work, which involved the proof-of-concept (POC) generating entire customised datasets within 15 minutes.
Visimo proposed R&D and Colorado State University current Professor Steven Simske said: “The work has potential security, privacy, access control, and authentication applications in areas like tracing, object recognition, biometrics, and scene analysis.
“Taking the concepts of red team/blue team in cybersecurity and cyber-physical security to GAN networks is a natural progression in the field of machine intelligence and will help the blue teams be able to stay ahead in the arms race of data manipulation.”
In July last year, Visimo was selected for Project Coeus that will deliver a collaboration tool for the US DoD.