A NEW APPROACH FOR AUDIO-VISUAL SEGMENTATION

Summary from the TLDR AI newsletter

This paper introduces the Explicit Conditional Multimodal Variational Auto-Encoder (ECMVAE) for better identification of sound sources in videos.