Two teams from our institute participated in the highly competitive DCASE Challenge and secured top positions in their respective tasks. The DCASE Challenge, which stands for "Detection and Classification of Acoustic Scenes and Events," is an international competition that pushes the boundaries of audio analysis and understanding.

Florian Schmid, Tobias Morocutti, Shahed Masoudian, Khaled Koutini, and Gerhard Widmer participated in Task 1: Low-Complexity Acoustic Scene Classification. This task focuses on developing algorithms capable of classifying audio scenes accurately, with a specific emphasis on low computational complexity. Our team successfully balanced model performance and computational complexity, earning them the first rank.

Simultaneously, Paul Primus, Khaled Koutini, and Gerhard Widmer participated in Task 6b: Language-Based Audio Retrieval. This task revolves around designing systems that can retrieve audio content based on textual descriptions, merging the domains of language processing and audio pattern recognition. Their submission, which utilized the proven PaSST model developed at the CP Institute, attained the highest position in the competition.

You can find more info on the two tasks here, opens an external URL in a new window and here, opens an external URL in a new window.