Audio transcription is a complicated problem in the field of music information retrieval. In this work, we attempt automatic transcription directly from audio waveforms using a uses a BLSTM network and a time-series MIDI representation.
Loading Thumbnail...