You can deploy this model and start earning money today!

stelee/
piano-transcription-onnx

Name: piano-transcription-onnx
Availability: InStock

Views

No views yet

Piano Transcription (high-resolution CRNN) — ONNX

ONNX export of the ByteDance / Qiuqiang Kong high-resolution piano transcription model (Kong et al. 2020, onset F1 96.7% on MAESTRO, MIT). Source checkpoint: Genius-Society/piano_trans.

Used by the megenta-jam PGM console for on-device piano stem → MIDI.

I/O

input audio: float32 [1, 160000] mono @ 16 kHz (10 s segment)
outputs: reg_onset, reg_offset, frame, velocity (each [1,1001,88]), plus pedal heads. Post-process with regression onset/offset decoding.

ORT note: load with graph optimizations disabled (ORT 1.20 MatMul+BN fusion mis-infers a rank-3 Gemm in this graph).

stelee/piano-transcription-onnx

Piano Transcription (high-resolution CRNN) — ONNX

I/O

stelee/
piano-transcription-onnx