Özkan Çayli

Drone detection and localization (DDL) aims to identify and track aerial targets using sensors. This is crucial for several potential applications, such as surveillance, security, and safety-critical monitoring. Recent advances in uncertainty-aware deep learning have made reliable DDL increasingly viable. However, achieving both accurate predictions and trustworthy uncertainty estimates remains challenging under low signal-to-noise ratios (SNRs) and when distinguishing valid targets from noise-only failure cases. In this paper, we introduce two complementary uncertainty-aware frameworks for acoustic DDL based on conformal prediction and evidential learning, which provide distribution-free coverage guarantees and model-based uncertainty decomposition, respectively. We further compare their localization performance against a heteroscedastic regression baseline. Our results demonstrate that while conformal prediction reliably guarantees statistical coverage, evidential learning excels in correlating uncertainty with prediction error. However, analysis of noise-only inputs reveals that distinguishing valid targets from failure cases remains a critical challenge, as models tend to retain high confidence even on out-of-distribution data. These findings highlight the necessity of combining uncertainty methods to build robust DDL systems.

Due to the widespread use of drones in an urban environment, drones present an increased risk to the safety of urban life. Reliable detection of drones becomes crucial for countering the hazard introduced by drones. However, drones are difficult to detect because of their size and customization. This paper introduces DDL, a dataset aimed at drone sound detection, classification, and localization via a specially constructed set of microphones. As a baseline, we propose a deep uncertainty-aware framework implementing Conformer for joint drone classification and localization. We employ heteroscedastic loss functions that jointly estimate means and variances for spatial localization to model prediction uncertainty. Experiments on the DDL dataset demonstrate a classification accuracy of 99.9% and a Euclidean distance mean absolute error (MAE) of approximately 16 meters. The uncertainty estimates are well-calibrated, with coverage closely matching the expected confidence intervals (68%, 95%, and 99.7%) as defined by the empirical rule, suggesting DDL as a benchmark dataset for audio-based drone localization.