The Radiological Anomaly Detection And Identification (RADAI) and REX (RADAI EXtended) projects are DOE NA22-funded efforts to produce high quality synthetic urban search datasets for the training and comparison of anomaly detection and identification algorithms. These datasets are the result of several years of NA22-funded research into the modeling of urban gamma-ray backgrounds, and an earlier version of the dataset was used in a public data competition on TopCoder (whose data is also available on BDC - login required).
After continuing to improve the realism of the simulation and increasing the number of source types, we are now releasing the latest version and making it available to researchers to use for the training and development of their own algorithms. This page describes how the data were produced and what they represent, how a researcher like you can download and use the data, and how you can participate in an ongoing leaderboard of the best results on the main testing dataset.
These data are inspired by public computer vision datasets like the ImageNet challenges, which have driven major developments in computer vision. With this open dataset and Leaderboard (login required), we are aiming to drive the development of radiation detection and identification algorithms for urban search.
To download the data, you first will need to register an account on the Berkeley Data Cloud (bdc.lbl.gov). If you do not see an appropriate organization to select during registration, select 'other' and write a message to admins requesting a new organization for your institution/research team. All academic and US national laboratory researchers' account requests will be approved, typically within 24 hours.
The data are being released under a CC-BY 4.0 license, meaning that you are free to share and adapt the data so long as attribution is given. Appropriate attribution is to cite ORNL/TM-2021/3265 (note we have a journal article in preparation that will replace this).
After using the developer and training datasets to improve your algorithm, you are welcome to upload your results (login required) on the testing dataset to the website for scoring. What you get out of scoring is a brief report on your algorithm’s performance on the testing set, broken down with some granularity into performance against different types of sources and false alarm rates, which could help you improve your approach. Your algorithm’s best submissions will also be shared on the website’s leaderboard.
Simulation details
References
Douglas E. Peplow, Daniel E. Archer, Andrew D. Nicholson, Nicholas J. Prins, Mark S. Bandstra, A. Chandler Jones, Brian J. Quiter, Abigael C. Nachtsheim, James M. Ghawaly, Jr., Threat Sources for Creating Synthetic Urban Search Data, ORNL/TM-2021/3265, Oak Ridge National Laboratory, Oak Ridge, Tennessee, February 2024
C.M. Anderson-Cook, D. Archer, M.S. Bandstra, J.C. Curtis, J.M. Ghawaly, T.H.Y. Joshi, K.L. Myers, A.D. Nicholson, and B.J. Quiter, ”Radiation Detection Data Competition Report”. United States: N. p., 2021. Web. https://doi:10.2172/1778748
A. D. Nicholson et al., "Multiagency Urban Search Experiment Detector and Algorithm Test Bed," in IEEE Transactions on Nuclear Science, vol. 64, no. 7, pp. 1689-1695, July 2017, https://doi.org/10.1109/TNS.2017.2677092
Swinney, M. W., Peplow, D. E., Patton, B. W., Nicholson, A. D., Archer, D. E., & Willis, M. J. (2018). A Methodology for Determining the Concentration of Naturally Occurring Radioactive Materials in an Urban Environment. Nuclear Technology, 203(3), 325–335. https://doi.org/10.1080/00295450.2018.1458558
Ghawaly, J.M., Nicholson, A.D., Peplow, D.E. et al. Data for training and testing radiation detection algorithms in an urban environment. Sci Data 7, 328 (2020). https://doi.org/10.1038/s41597-020-00672-2
A. D. Nicholson, D. E. Peplow, J. M. Ghawaly, M. J. Willis and D. E. Archer, "Generation of Synthetic Data for a Radiation Detection Algorithm Competition," in IEEE Transactions on Nuclear Science, vol. 67, no. 8, pp. 1968-1975, Aug. 2020. https://doi.org/10.1109/TNS.2020.3001754
Peplow, Douglas, Archer, Daniel, Ghawaly, James, Joshi, Tenzing H.Y., Bandstra, Mark S., Quiter, Brian, and Nachtsheim, Abigael C. Detection Algorithm Virtual Testbed for Urban Search with SCALE. United States: N. p., 2022. Web. https://doi.org/10.13182/ICRSRPSD22-39450 https://www.osti.gov/biblio/1897851
Peplow, Douglas, Archer, Daniel, Ghawaly, James, Joshi, Tenzing H.Y., Bandstra, Mark S., and Quiter, Brian. Threat Sources for Detection Algorithm Testing Developed with SCALE. United States: N. p., 2022. Web. https://doi.org/10.13182/ICRSRPSD22-39449 https://www.osti.gov/biblio/1897855
Swinney, Mathew W.; Peplow, Douglas E.; Nicholson, Andrew D.; and Patton, Bruce W. NORM Concentration Determination in Common Materials in an Urban Environment. United States: N. p., 2016. Web. https://www.osti.gov/biblio/22991969
“Berkeley Lab Researchers Develop Platform for Hosting Science Data Analytics Competitions,” April 16 2018, https://crd.lbl.gov/news-and-publications/news/2018/berkeley-lab-researchers-develop-platform-for-hosting-science-data-analytics-competitions/