Objects that Sound

arXiv 2017

In this paper our objectives are, first, networks that can embed audio and visual inputs into a common space that is suitable for cross-modal retrieval; and second, a network that can localize the object that sounds in an image, given the audio signal. We achieve both these objectives by training from unlabelled video using only audio-visual correspondence (AVC) as the objective function. This is a form of cross-modal self-supervision from video. To this end, we design new network architectures that can be trained using the AVC task for these functionalities: for cross-modal retrieval, and for localizing the source of a sound in an image. We make the following contributions: (i) show that audio and visual embedding can be learnt that enable both within-mode (e.g. audio-to-audio) and between-mode retrieval; (ii) explore various architectures for the AVC task, including those for the visual stream that ingest a single image, or multiple images, or a single image and multi-frame optical flow; (iii) show that the semantic object that sounds within an image can be localized (using only the sound, no motion or flow information); and (iv) give a cautionary tale in how to avoid undesirable shortcuts in the data preparation.

View publication Download publication

Latest research news

View the blog