Synthetik propose to leverage a new generation of high-resolution, low-cost (e.g. less than $1,000, 25-100x less than currently available systems) hyperspectral cameras in conjunction with state-of-the-art machine learning to rapidly image, process and predict locations where residues of interest are present to deliver a targeted, relevant and more dense samples to the trace detector.During Phase I, Synthetik plan to adapt their Spektra high-resolution, low-cost hyperspectral camera system (developed as part of a NAA-funded industrial and food safety techNlogy development project) for application to this DHS requirement.Training of state-of-the-art machine learning models suitable forhyperspectral data will allow discrimination between background material and areas of interest containing residues and this will provide targeting information to the screener. This concept will be demonstrated to DHS at the close of Phase I.Hyperspectral imaging provides a modality that is well-suited to this task. The techNlogy is a natural, multi-dimensional spatial extension of standard spectroscopy techniques, whereby each pixel of the captured image represents a complete spectral signature at a given spectral resolution (e.g. wavelength). In this way, a hyperspectral image can be viewed as a data cube which contains both spatial information and spectral (e.g. wavelength, intensity) information for each captured image.Critically, when paired with modern deep learning-based semantic segmentation models, the captured image can be used to identify and classify materials, including traces of explosive or illicit drugs, and then show the operator where these materials are on the article of interest