1 paper · 1 filter
Emiliano Acevedo, Martín Rocamora, Magdalena Fuentes
Audio-text models are widely used in zero-shot environmental sound classification as they alleviate the need for annotated data. However, we show that their performance severely dr…