"a rotary-wing drone acoustic dataset (RWDA dataset)"

"The RWDA dataset is developed to facilitate research on robust drone detection within complex acoustic environments characterized by significant domain shifts. This collection includes audio samples from two distinct categories which represent drone signals and various natural environmental noises recorded in open field settings. The drone category contains recordings of the DJI Air 3S and the DJI Mini 4K performing diverse flight manoeuvres such as acceleration, deceleration, constant speed flight, and hovering at altitudes ranging from 5 to 120 metres. Data acquisition was executed via two 32-channel arrays with a sampling rate of 48kHz and a supplementary 64-channel array operating at 150kHz. To evaluate the generalization capabilities of detection methods, the dataset is partitioned such that urban samples are designated for training while mountainous samples are utilized for validation and testing. This configuration introduces substantial discrepancies in both the acoustic background and the signal-to-noise ratio to simulate challenging real-world deployment scenarios."