The Department of Homeland Security (DHS) gives notice of the availability of the “Department of Homeland Security’s Person Screening Algorithm Challenge Prize Competition and rules.” The DHS Science and Technology Directorate ( S&T) Homeland Security Advanced Research Projects Agency (HSARPA) Explosives Division (EXD) and the Transportation Security Administration (TSA) (Competition Sponsor) are seeking new automated detection algorithms from individuals and entities that improve the speed, accuracy, and detection of small threat objects and other prohibited items during the airport passenger screening process.
Algorithms developed from this Competition, or through further research and development under a limited intellectual property use agreement, have the potential to improve the speed, the detection of prohibited items, and the accuracy of Advanced Imaging Technology (AIT) scanners. A comprehensive set of new automated detection algorithms have the potential to be integrated into the latest screening equipment.
In 2012, Congress required all person screening technology to have automated target recognition. DHS S&T and TSA are developing better screening technologies and want to promote the development of the best automated detection algorithms for TSA screening equipment that have the potential to advance the DHS mission.
Traditional algorithms used in automated screening equipment are company proprietary and must be adapted to unique detection hardware geometrics. When a single change in screening requirements or threat detection is required, each Original Equipment Manufacturer (OEM) must adapt their proprietary algorithm. Instead of contracting with specific OEM’s on a one-on-one basis, DHS intends to use this Prize Competition to explore the possibility of discovering a comprehensive or set of automated detection algorithms that can improve the speed, detection of prohibited items, and accuracy in both OEM and government developed screening equipment. The goal of the Prize Competition is to reach beyond the traditional screening equipment manufacturers to entities who are exclusively skilled at algorithm development. This provides an opportunity to discover new talent that may otherwise be overlooked by DHS or current technology providers.
This Prize Competition consists of two stages. The first stage will give competitors six months to train and build automated threat detection algorithms using a dataset of hundreds of laboratory study volunteer mock passenger research images in a variety of formats. The second stage will evaluate those models against a holdout set of images for the purpose of validating algorithm performance.
The total cash prize payout for this Competition is up to $1,500,000 (USD).
Overview: The Competition will be administered and run on Kaggle.com. Submissions, in the form of predictions on the dataset, can be made to the Kaggle platform during the advertised competition submission period. The accuracy of those submissions will be evaluated using the multi-class logarithmic loss metric. The top eight ranked teams at the end of the second stage will be eligible to receive prize money, distributed as specified under “Prizes.”
Privacy: Personal information provided by entrants (Solvers) on the nomination form through the Prize Competition website will be used to contact selected finalists. Information is not collected for commercial marketing. Winners are permitted to cite that they won this Competition after it has been officially announced by the Department of Homeland Security. The names, cities, and states of selected winner or entity will be made available in promotional materials and at recognition events.
Awards:- The top eight ranked teams at the end of the second stage Competition will be eligible to receive a cash prize, distributed as follows:
1st Place: $500,000
2nd Place: $300,000
3rd Place: $200,000
4th – 8th Place: $100,000 each