FUSS

该数据集名为FUSS，被用作训练能够分离各种任务中音频源的上游数据集。此外，在FUSS上进行微调可以提高源计数准确性和SI-SDRi，这表明模型的迁移性得到了增强。该任务的目的是实现音频源分离。This dataset, named FUSS, serves as an upstream dataset for training models capable of audio source separation across diverse tasks. Furthermore, fine-tuning on FUSS can improve source count accuracy and SI-SDRi, indicating enhanced model transferability. The objective of this task is audio source separation.