BatVision

BatVision数据集由巴黎高等矿业学院和密歇根大学合作创建，旨在从机器人视角研究视听数据。数据集包含大规模的音频-视觉数据，通过装备有扬声器和双耳麦克风的机器人，在历史校园和现代办公楼中模拟蝙蝠行为，发出啁啾声并记录其回声。这些回声携带着丰富的场景信息，包括物体、材料和布局。数据集不仅包括回声，还提供相机图像和深度图，旨在解决声音与空间交互的基本问题，以及如何利用声音进行机器人导航和场景理解。

The BatVision dataset was co-developed by École des Mines de Paris and the University of Michigan, with the goal of investigating audio-visual data from a robotic perspective. It comprises large-scale audio-visual data collected via a robot fitted with speakers and binaural microphones. The robot simulates bat behavior in both historic campus buildings and modern office structures, emitting chirps and recording their echoes. These echoes contain rich scene-related information, including objects, material properties, and spatial layouts. Besides the echo data, the dataset also provides camera images and depth maps. The core objectives of this dataset are to address fundamental problems of sound-space interaction, as well as to explore how sound can be utilized for robotic navigation and scene understanding.