Cignal creates new software systems and capabilities using high-fidelity synthetic data that facilitate the rapid training and deployment of advanced vision systems.
“Having a source for computer-generated and annotated training images will significantly reduce the time and expense for fielding new machine learning models for more efficient screening and in response to new threats” said Karl Harris, S&T Screening at Speed Program Manager. “Generating large amounts of training data is typically a very labor-intensive, manual process, with some object recognition models requiring tens of thousands of labeled images per object class.”
Cignal project proposes to enhance its working prototype and training workflow product, Cignal Workbench, to generate high-fidelity synthetic training data for Advanced Technology (AT) X-ray inspection systems and synthetic volumetric data for Computed Tomography (CT) applications. The result will be a high-volume data source for labeled baggage for seamless, unsupervised, and continuous AI model training.
“The system and architecture defined in this application will be capable of automated, unsupervised training on a billion baggage images,” said Melissa Oh, SVIP Managing Director. “Access to labeled AT and CT training data has the potential to create advanced detection systems with superior capabilities.”
SVIP is one of the programs and tools available for S&T to fund innovation and engage with private sector partners to advance homeland security solutions. Companies participating in SVIP are eligible for up to $800,000 of non-dilutive funding over four phases to develop and adapt commercial technologies for homeland security use cases.
For more information about S&T’s innovation programs and tools, visit https://www.dhs.gov/science-and-technology/business-opportunities.
Original post https://alertarticles.info