1 paper
Ruigang Li, Yongxu Zhu
Existing multi-timbre transcription models struggle with generalization beyond pre-trained instruments, rigid source-count constraints, and high computational demands that hinder d…