Multitask learning for instrument activation aware music source separation
This work addresses music source separation for music information retrieval, offering incremental gains by integrating related tasks.
The paper tackled music source separation by proposing a multitask model that incorporates instrument activation information, achieving performance improvements over the baseline Open-Unmix model on a combined dataset of Mixing Secrets and MedleyDB while maintaining comparable results on MUSDB.
Music source separation is a core task in music information retrieval which has seen a dramatic improvement in the past years. Nevertheless, most of the existing systems focus exclusively on the problem of source separation itself and ignore the utilization of other~---possibly related---~MIR tasks which could lead to additional quality gains. In this work, we propose a novel multitask structure to investigate using instrument activation information to improve source separation performance. Furthermore, we investigate our system on six independent instruments, a more realistic scenario than the three instruments included in the widely-used MUSDB dataset, by leveraging a combination of the MedleyDB and Mixing Secrets datasets. The results show that our proposed multitask model outperforms the baseline Open-Unmix model on the mixture of Mixing Secrets and MedleyDB dataset while maintaining comparable performance on the MUSDB dataset.