%0 Conference Proceedings %B 14th European Conference on Computer Vision %D 2016 %T Lecture Notes in Computer ScienceComputer Vision – ECCV 2016Ambient Sound Provides Supervision for Visual Learning %A Owens, Andrew %A Isola, P. %A Josh H. McDermott %A William T. Freeman %A Torralba, Antonio %K convolutional networks %K Sound %K unsupervised learning %X

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.

%B 14th European Conference on Computer Vision %C Cham %P 801 - 816 %8 10/2016 %@ 978-3-319-46447-3 %G eng %U http://link.springer.com/10.1007/978-3-319-46448-0 %R 10.1007/978-3-319-46448-010.1007/978-3-319-46448-0_48