BirdWatch: Acoustic Drone Detection

The Birdwatch project addresses the growing threat of small, stealthy drones, particularly first-person-view (FPV) and fly-by-wire platforms, in modern conflict zones by developing a portable, decentralized acoustic detection system tailored for frontline military personnel. Unlike existing centralized acoustic networks, which excel at detecting large drones but lack tactical mobility, Birdwatch focuses on real-time, low-latency detection of small drones using lightweight hardware, enabling dismounted soldiers and lightly armored vehicles to evade or counter threats with critical seconds of warning. The system leverages machine learning models optimized for edge deployment, combining binary classification with temporal modeling to infer drone movement, direction, and proximity, key gaps in current solutions. By integrating multimodal data (audio + video) and Time-Difference of Arrival (TDoA) processing, the project aims to surpass traditional spectrogram-based approaches, offering a scalable, noise-resilient framework adaptable to diverse environments (urban, rural, coastal) and operational constraints.