Multiple Sound Sources Localization from Coarse to Fine
This addresses the challenge of multi-source sound localization in complex scenes for applications like video analysis, though it is incremental as it builds on existing audiovisual methods.
The paper tackles the problem of visually localizing multiple sound sources in unconstrained videos without pairwise sound-object annotations by developing a two-stage audiovisual learning framework that disentangles and aligns audio and visual features in a coarse-to-fine manner, achieving state-of-the-art results on a public dataset and comparable performance in sound separation.
How to visually localize multiple sound sources in unconstrained videos is a formidable problem, especially when lack of the pairwise sound-object annotations. To solve this problem, we develop a two-stage audiovisual learning framework that disentangles audio and visual representations of different categories from complex scenes, then performs cross-modal feature alignment in a coarse-to-fine manner. Our model achieves state-of-the-art results on public dataset of localization, as well as considerable performance on multi-source sound localization in complex scenes. We then employ the localization results for sound separation and obtain comparable performance to existing methods. These outcomes demonstrate our model's ability in effectively aligning sounds with specific visual sources. Code is available at https://github.com/shvdiwnkozbw/Multi-Source-Sound-Localization