OVAD - Occluded Vehicle Acoustic Detection dataset

Acoustic multi-microphone dataset (plus video) from the paper "Hearing What You Cannot See: Acoustic Vehicle Detection Around Corners" (IEEE Robotics and Automation Letters 2021, see DOI: 10.1109/LRA.2021.3062254 and arxiv). For a short and intuitive introduction of our main ideas and online prediction results, we recommend to watch our supplementary video on Youtube: Hearing What You Cannot See.<br>We present a data-driven acoustic detection method that can detect an approaching vehicle before it enters line-of-sight, e.g. while hidden behind a blind corner, in real-world outdoor conditions. The code can be found at github.com/tudelft-iv/occluded_vehicle_acoustic_detection .<br>As part of this publication, we provide a novel audio-visual dataset (OVAD - occluded vehicle acoustic detection) collected in outdoor urban environments with a 56-microphone array mounted on our research vehicle.

本数据集源自2021年发表于《IEEE机器人与自动化快报》（IEEE Robotics and Automation Letters）的论文《Hearing What You Cannot See: Acoustic Vehicle Detection Around Corners》，其DOI为10.1109/LRA.2021.3062254，预印本可在arxiv平台获取，包含多麦克风声学数据与配套视频。若需快速直观了解本研究的核心思路与在线预测效果，建议观看我们在Youtube发布的补充视频《Hearing What You Cannot See》。本研究提出了一种数据驱动的声学检测方法，可在真实户外场景中，于接近的车辆进入视距范围内（例如仍被盲角遮挡时）提前探测到该车辆。相关代码可访问github.com/tudelft-iv/occluded_vehicle_acoustic_detection 获取。作为该论文的附属配套成果，我们提供了一套全新的音视频数据集（OVAD——遮挡车辆声学检测数据集，Occluded Vehicle Acoustic Detection），该数据集采集自户外城市环境，采集设备为安装在我方研究车辆上的56通道麦克风阵列。