Abstract
Music classification, with a wide range of applications, is one of the mostprominent tasks in music information retrieval. To address the absence ofcomprehensive datasets and high-performing methods in the classification ofmainstage dance music, this work introduces a novel benchmark comprising a newdataset and a baseline. Our dataset extends the number of sub-genres to covermost recent mainstage live sets by top DJs worldwide in music festivals. Acontinuous soft labeling approach is employed to account for tracks that spanmultiple sub-genres, preserving the inherent sophistication. For the baseline,we developed deep learning models that outperform current state-of-the-artmultimodel language models, which struggle to identify house music sub-genres,emphasizing the need for specialized models trained on fine-grained datasets.Our benchmark is applicable to serve for application scenarios such as musicrecommendation, DJ set curation, and interactive multimedia, where we alsoprovide video demos. Our code is on\url{https://anonymous.4open.science/r/Mainstage-EDM-Benchmark/}.