Lecture Notes in Computer ScienceComputer Vision – ECCV 2016Ambient Sound Provides Supervision for Visual Learning

TitleLecture Notes in Computer ScienceComputer Vision – ECCV 2016Ambient Sound Provides Supervision for Visual Learning
Publication TypeConference Proceedings
Year of Publication2016
AuthorsOwens, A, Isola, P, McDermott, JH, Freeman, WT, Torralba, A
Conference Name14th European Conference on Computer Vision
Pagination801 - 816
Date Published10/2016
Conference LocationCham
ISBN Number978-3-319-46447-3
ISBN0302-9743
Keywordsconvolutional networks, Sound, unsupervised learning
Abstract

The sound of crashing waves, the roar of fast-moving cars – sound conveys important information about the objects in our surroundings. In this work, we show that ambient sounds can be used as a supervisory signal for learning visual models. To demonstrate this, we train a convolutional neural network to predict a statistical summary of the sound associated with a video frame. We show that, through this process, the network learns a representation that conveys information about objects and scenes. We evaluate this representation on several recognition tasks, finding that its performance is comparable to that of other state-of-the-art unsupervised learning methods. Finally, we show through visualizations that the network learns units that are selective to objects that are often associated with characteristic sounds.

URLhttp://link.springer.com/10.1007/978-3-319-46448-0
DOI10.1007/978-3-319-46448-010.1007/978-3-319-46448-0_48

Research Area: 

CBMM Relationship: 

  • CBMM Funded