The study titled ‘Multi-label Open-set Audio Classification’ by Sripathi Sridhar and Mark Cartwright discusses the limitations of current audio classification models and the need for open-set audio classification techniques. These models are designed to detect sound events from unknown classes, which is critical in real-world scenarios where sounds often overlap and can come from unknown or new sources. The study establishes the problem of multi-label open-set audio classification by creating a dataset with varying unknown class distributions and evaluating baseline approaches.
Publication date: 25 Oct 2023
Project Page: N/A
Paper: https://arxiv.org/pdf/2310.13759
