Synthetic Data Validation

Tender description

The Transport Security, Resilience and Response (TSRR) within DfT, leads on national security matters, ranging from counter terrorism and cyber security to planning for and responding to natural hazards or civil contingencies. We deliver expertise to support security and resilience policy teams whilst also providing the department's 24/7 response for all security or civil contingency incidents. The team works with security and intelligence partners from across government to mitigate risk to the UK's transport sector. For the development of algorithms for the automated detection of security threats there is a need for data sets of both the threats to be found and of innocuous items to ensure that the algorithm performance does not impede on the operational needs of airports. With the advent of machine and deep learning algorithms there is a requirement for large volumes of data, this, combined with the fact that some of the materials that regulators wish algorithms to identify are hard to handle in large quantities meaning that the creation of data is slow or inadequate to meet the demands of development and deployment of algorithms to meet all stakeholder needs. A potential solution to this problem would be to use artificially created data and use this to both learn and validate algorithm performance but before using this approach it would be pertinent to understand how accurate synthetic data is compared to data collected on a real machine. The purpose of this contract would be to develop a means by which synthetically created data could be created and validated as to be identical and indistinguishable from data created by scanning real objects in security systems. If validated DfT would then look to have created a synthetic data set of bags with threats created by an identical means as that validated to be distributed to vendors for algorithm development and validation. The scope of this program is for the validation of synthetically created images for x-ray computer tomography systems for the purpose of being used to develop algorithms for the automated detection of prohibited items. Please access the opportunity via the Jaggaer E-Sourcing Portal; tender_1748 - Synthetic Data Validation