Note-level Singing Voice Transcription on Mandarin datasets M4Singer and D1 (noisy)
93.8COn (F)ROSVOT
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ROSVOTconfiguration=final (ours)2024.05 | 93.8 | 94.4 | 77 | 97.1 | 87.4 | |
| ROSVOTablation=w/ Lw2024.05 | 93.5 | 92.9 | 77 | 96.2 | 87.2 | |
| ROSVOTvariant=conv2024.05 | 91.5 | 92.6 | 70.8 | 96.8 | 84.1 | |
| ROSVOTablation=w/o wbd2024.05 | 91.1 | 91.2 | 69.9 | 95.1 | 83.4 | |
| ROSVOTablation=w/o noise2024.05 | 90.9 | 91.5 | 70.1 | 95.2 | 83.1 | |
| ROSVOTvariant=conformer2024.05 | 90.6 | 90.8 | 69.8 | 95.3 | 83.1 | |
| Yong et al.year=20232024.05 | 88.5 | 89.7 | 62.1 | 86.4 | 80.6 | |
| MusicYOLO2024.05 | 79.7 | 76.5 | 51.5 | 78.6 | 78.9 | |
| VOCANO2024.05 | 64.7 | 66.1 | 43.4 | 71.9 | 59.8 | |
| TONY2024.05 | 49.2 | 47 | 28.4 | 46.6 | 45.2 |