Structure from Silence: Learning Scene Structure from Ambient Sound
This addresses the problem of enhancing multimodal AI perception for robotics or AR/VR by leveraging ambient sound, though it is incremental as it builds on existing multimodal self-supervision methods.
The paper tackled the problem of whether ambient sounds can convey 3D scene structure, and found that models trained on audio alone can estimate distances to walls, with results suggesting ambient sound provides useful information for multimodal learning.
From whirling ceiling fans to ticking clocks, the sounds that we hear subtly vary as we move through a scene. We ask whether these ambient sounds convey information about 3D scene structure and, if so, whether they provide a useful learning signal for multimodal models. To study this, we collect a dataset of paired audio and RGB-D recordings from a variety of quiet indoor scenes. We then train models that estimate the distance to nearby walls, given only audio as input. We also use these recordings to learn multimodal representations through self-supervision, by training a network to associate images with their corresponding sounds. These results suggest that ambient sound conveys a surprising amount of information about scene structure, and that it is a useful signal for learning multimodal features.