This project addressed the challenge of timely threat detection in crowded public places by developing human-in-the-loop gun detection solutions. The proposed solution intended to improve emergency response times and resource allocation by integrating AI-driven detection with real-time visualization capabilities. The project aimed to improve the efficiency of law enforcement and emergency services by providing real-time threat detection (particularly for detecting guns in videos and marking the corresponding videos as anomaly) and response recommendations through a visualization dashboard.
The project utilized AI-based computer vision algorithms, including deep learning models such as convolutional neural networks (CNNs) and long short-term memory networks (LSTMs). These algorithms were tailored to detect guns in live-stream video data from surveillance cameras in public spaces. The project intended to apply its findings through a real-time visualization dashboard, including a multi-layer security and authentication system with different access controls for different roles, such as administrator, supervisor, and operator. The dashboard provided an interactive map to help law enforcement quickly identify potential threats and provide information to support effective decision-making in a crisis.
Existing technology is subject to human error and consumes a significant amount of time for processing, which hinders real-time detection capabilities. Researchers aimed to reduce human error by using an AI-based gun detection as well as a real-time visualization dashboard. Integrating an interactive map within the dashboard intended to provide a more responsive and effective decision-support tool, enhancing decision-making processes in a crisis and improving overall public safety. Further, researchers had planned to utilize differential privacy techniques to better preserve privacy.
This project began in 2024 and identified the following resources in its first year: a platform for threat detection in crowded places; datasets to leverage in support of this work; and existing tools to inform how to move forward with the research; Started planning for a demo on showing preliminary gun detection methods using the developed dashboard.
The project intended to leverage insights from other DHS-funded initiatives, including the ADMIRE Center and the Criminal Investigations and Network Analysis Center (CINA). Both the SENTRY Practitioner Advisory Board and the Industry Advisory Board provided recommendations and insights to tailor the decision support tool to specific end-users’ needs. These collaborations enhanced the project’s applicability to end-users and facilitate knowledge sharing across related security and defense projects.
