CANDLE: Privacy-preserving concealed weapon detection with AI thermal fusion on mobile edge devices.

CANDLE is selected during the dAIEDGE 2nd Open Call. The project addresses the need for real-time concealed weapon detection while preserving privacy with automatic face-blurring, aligning with dAIEDGE’s scope of supporting Europe’s leadership in responsible AI innovation, strengthening the collective AI ecosystem, and safeguarding privacy. The project develops a privacy-preserving thermal vision pipeline for concealed weapon detection in real time on widely available smartphones by leveraging AI optimization techniques. CANDLE minimizes invasive scanning and centralized data storage through a portable detection solution for law enforcement and first responders in EU states, adhering to the EU’s ethical AI standards with robust on-device privacy measures. The integration with the dAIEDGE VLab, enables the sharing of pre-processed thermal datasets and benchmarking results through VLab’s remote benchmarking suite, fostering collaborative experimentation, reproducibility, and accelerated innovation in European edge AI.

The offered solution is affordable and widely deployable in real-world scenarios, from crowded public events to local law enforcement applications, enhancing situational awareness for users (e.g., police) with all faces remaining blurred, ensuring reliable threat detection without exposing personal data, in compliance with GDPR. CANDLE performs on-device thermal inference at the edge, eliminating the need for high-end hardware or cloud resources, lowering costs and latency. The solution generates a significant environmental impact, as it reduces the need for energy-intensive data transmissions and centralized cloud processing, utilizing edge AI for performing detection directly on mobile devices. Socially, CANDLE enhances public safety by equipping law enforcement and first responders with an ethically operated detection toolcnthat preserves privacy. CANDLE’s robustness will be tested through extensive trials in controlled and simulated public safety scenarios.

For more information, you can visit the dAIEDGE website.