A 16-Channel Acoustic Dataset for Passive Ground-Based Monitoring of Small Drones

Synchronized 16-channel acoustic recordings for passive ground-based monitoring of small UAVs, collected with an SMD-CR16 planar cross array: 200 hovering DJI Flip recordings, 28 field background recordings, and 45 controlled re-recordings of non-UAV sounds. The 273 physical recordings comprise 4,368 mono WAV files (48 kHz, 16-bit PCM), totaling 7.221 hours. The UAV subset spans four azimuths, four heights (5–30 m), six horizontal distances (5–70 m), and quiet outdoor and road-adjacent urban environments. Microphone coordinates, recording-level JSON metadata, CSV indexes, replay source provenance and integrity hashes accompany the audio. Controlled replay source excerpts were selected from FSD50K entries licensed under CC0 1.0. The archive includes validation code and results, benchmark implementations, original feature and CNN support modules, configurations, recording-level folds and predictions, training histories, summary tables, plotting data, and a portable reproduction entry point. It covers detection, negative-sample composition, azimuth classification and supplementary distance regression. Trained weights and feature caches are regenerated by the supplied code. No models were retrained during packaging.