DuET: Note--Frame Dual Encoding for Piano Performance-to-Score Transcription

Model

DuET transcribes piano performance MIDI into structured musical scores with a note encoder, a frame encoder, and a note-aligned autoregressive decoder. The model jointly learns score prediction and frame-level beat and downbeat tracking, using TimeRoPE and TPRoPE to represent relative time and pitch relationships.

GitHub: code, installation, and usage

Paper: DuET: Note--Frame Dual Encoding for Piano Performance-to-Score Transcription