MAESTRO
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Automatic Piano Transcription | MAESTRO v3.0.0 (test) | Note F1 Score (w/ Offset)93.48 | 19 | |
| Integer Regression | MAESTRO (test) | RMSE32 | 13 | |
| Audio Inpainting | MAESTRO | ODG (PEA-Q)-3.318 | 8 | |
| Music Transcription | MAESTRO HEAR challenge | Onset FMS0.391 | 8 | |
| Audio Watermarking | MAESTRO | Detectability Accuracy100 | 7 | |
| Unconditional Symbolic Music Generation | Maestro | Used Pitch Coverage0.905 | 7 | |
| Audio-MIDI Retrieval | MAESTRO long-horizon (train) | Best S (%)83.8 | 6 | |
| Audio Generation | MAESTRO (test) | FAD (unconditional)0.131 | 6 | |
| Piano Transcription | MAESTRO (test) | Frame Precision93.1 | 6 | |
| Sustain Pedal Transcription | Maestro v3.0.0 (test) | Activation Precision96.69 | 5 | |
| Music transcription | MAESTRO | Onset FMS0.391 | 5 | |
| Pedal Transcription | MAESTRO (test) | Frame Precision94.41 | 5 | |
| Retrieval | MAESTRO | Mean S Score (%)84 | 4 | |
| Automatic Music Transcription | MAESTRO (test) | Note Onset F189.7 | 4 | |
| Audio Inpainting | MAESTRO (test) | MOS4.12 | 4 | |
| Music Transcription | MAESTRO | Frame F188 | 4 | |
| Unconditional Music Generation | MAESTRO | Selection Rate95.8 | 4 | |
| Blind Audio Enhancement | MAESTRO | FAD (VGG)0.84 | 3 | |
| Error Detection | MAESTRO-E | Correct (F1)94.4 | 3 | |
| Soft Pedal Transcription | Maestro v3.0.0 (test) | Activation Precision86.42 | 2 | |
| Symbolic music generation | MAESTRO v1 (val) | Negative Log-Likelihood1.82 | 2 | |
| Symbolic music generation | MAESTRO v3 (val) | NLL1.9 | 1 | |
| Symbolic music generation | MAESTRO v3 (test) | Negative Log-Likelihood1.91 | 1 | |
| Symbolic music generation | MAESTRO v1 (test) | Negative Log-Likelihood1.82 | 1 |
Showing 24 of 24 rows