A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is
In the field of music information retrieval, the task of simultaneously identifying the presence or absence of multiple musical instruments in a polyphonic recording remains a hard problem. Previous works have seen some success in improving instrument classification by applying temporal attention in a multi-instance multi-label setting, while another series of work has also suggested the role of pitch and timbre in improving instrument recognition performance. In this project, we furtherarXiv:2006.09640v2 fatcat:cwtyxoszdjfungn3hx7cemnao4