The Audio-Visual BatVision Dataset for Research on Sight and Sound

The BatVision dataset contains large scale audio-visual data from a robots’ perspective. For its creation, a robot traversed corridors, offices, lecture halls and driveways at a historic campus and modern office building like a bat, emitting chirping sounds with a speaker. A binaural microphone recorded their echoes which carry rich scene information of objects, materials and layout. This dataset will help investigate fundamental questions on how sound interacts with spaces, how it can be harnessed for robotic navigation and what in general can be understood about a scene from how it sounds.