In October 2025 a one-day community workshop was convened to explore the vision, scope, and feasibility of establishing a Centre for AI in Applied Nuclear Physics, with relevance to both civil and defence domains. The event brought together researchers, industry partners, and government stakeholders for a series of interactive sessions, keynote presentations, and a community poster...
Global geo-political and economic instabilities and uncertainties mean that our security services are under increasing pressure to respond to suspected and genuine incidents under time pressures and with limited resources. These conditions impose challenges on the equipment used and the underpinning technologies, in order to ensure ease of use, robustness under increasingly under...
Reliable UXO detection remains one of the most persistent sensing challenges due to the diversity of threats and the variability of real-world environments. This study explores the use of synthetic data to train object detection models for landmine detection. The YOLOV8n model trained on n=1000 real images achieved a mAP@50 of 0.983, a precision of 0.966 and a recall of 0.961. The best...
The methodology for generation of the SIGMA hybrid datasets is described from identification of candidate regions where threats can be injected, through template generation to subsampling into background data. The proposed format for data is also discussed, comments welcome
Over the last two years, a number of datasets have been made available to the UK academic community from ~100 SIGMA detectors deployed in London during 2017-2018. These datasets contain real threats and background, however, lack ground truth makes detailed analysis of detection probability and ID challenging. A new dataset with injected threats (known isotope, activity, speed) into regions...
