UaVirBASE: A Public-Access Unmanned Aerial Vehicle Sound Source Localization Dataset

UaVirBASE is a publicly available dataset designed for the sound source localization (SSL) of unmanned aerial vehicles (UAVs) using acoustic signals. It contains synchronized multi-microphone audio recordings collected under controlled experimental conditions, featuring a wide range of UAV positions and orientations with respect to a fixed microphone array. The dataset includes variations in: Distance and altitude from the array Azimuth angles UAV orientations (front, back, left, right) These variations are critical for training robust machine learning and deep learning models for SSL in real-world UAV scenarios. Additionally, this dataset is accompanied by detailed metadata and annotation files, as well as documentation outlining the recording setup, hardware specifications, and data acquisition pipeline. We provide a baseline deep neural network (DNN) model trained on UaVirBASE with performance metrics including: Mean Absolute Error (MAE) of 0.5 meters for distance/height ~1 degree azimuth error <10 degrees for side (orientation) error UaVirBASE is intended to support reproducible research and development in acoustic-based SSL, UAV surveillance, and intelligent sensing systems. This dataset fills a gap in the literature by offering a standardized benchmark specifically tailored to UAV audio-based localization tasks.Jekateryńczuk, G.; Szadkowski, R.; Piotrowski, Z. UaVirBASE: A Public-Access Unmanned Aerial Vehicle Sound Source Localization Dataset. Appl. Sci. 2025, 15, 5378. https://doi.org/10.3390/app15105378UaVirBASE是一款专为利用声学信号实现无人机（unmanned aerial vehicles, UAVs）声源定位（sound source localization, SSL）任务开发的公开可用数据集。该数据集包含在受控实验条件下采集的同步多麦克风音频录音，涵盖了相对于固定麦克风阵列的多种无人机位置与朝向场景。 数据集涵盖以下变化维度： - 与麦克风阵列的距离及飞行高度 - 方位角 - 无人机朝向（前、后、左、右） 上述变化维度对于训练适配真实无人机场景下声源定位任务的鲁棒机器学习与深度学习模型至关重要。 此外，本数据集配套提供了详细的元数据与标注文件，以及涵盖录音设置、硬件规格与数据采集流程的完整文档。我们还提供了基于UaVirBASE训练的基线深度神经网络（deep neural network, DNN）模型，其性能指标如下： - 距离与高度的平均绝对误差（Mean Absolute Error, MAE）为0.5米 - 方位角误差约为1度 - 朝向侧偏误差小于10度 UaVirBASE旨在支持基于声学的声源定位、无人机监测与智能感知系统领域的可复现研究与开发工作。本数据集填补了现有学术文献的空白，提供了专为无人机音频定位任务定制的标准化基准测试集。 Jekateryńczuk, G.; Szadkowski, R.; Piotrowski, Z. UaVirBASE: A Public-Access Unmanned Aerial Vehicle Sound Source Localization Dataset. Appl. Sci. 2025, 15, 5378. https://doi.org/10.3390/app15105378