Visually Indicated Sounds
This work addresses the challenge of understanding physical interactions in visual scenes for applications in robotics, virtual reality, and multimedia, representing a novel domain-specific advancement.
The paper tackles the problem of predicting sounds from silent videos of objects being struck, using a recurrent neural network to predict sound features and an example-based synthesis procedure to generate waveforms. The model's synthesized sounds were realistic enough to deceive participants in a psychophysical experiment and conveyed information about material properties and physical interactions.
Objects make distinctive sounds when they are hit or scratched. These sounds reveal aspects of an object's material properties, as well as the actions that produced them. In this paper, we propose the task of predicting what sound an object makes when struck as a way of studying physical interactions within a visual scene. We present an algorithm that synthesizes sound from silent videos of people hitting and scratching objects with a drumstick. This algorithm uses a recurrent neural network to predict sound features from videos and then produces a waveform from these features with an example-based synthesis procedure. We show that the sounds predicted by our model are realistic enough to fool participants in a "real or fake" psychophysical experiment, and that they convey significant information about material properties and physical interactions.