Note-level Singing Voice Transcription on Mandarin datasets M4Singer and D1 (clean)
94COn (F)ROSVOT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ROSVOTconfiguration=final (ours)2024.05 | 94 | 94.5 | 77.4 | 97 | 87.6 | |
| ROSVOTablation=w/o noise2024.05 | 93.8 | 94.2 | 76.4 | 97.1 | 87.1 | |
| ROSVOTablation=w/ Lw2024.05 | 93.3 | 93.2 | 77.1 | 96.5 | 87.5 | |
| ROSVOTvariant=conformer2024.05 | 92.1 | 91.8 | 70.3 | 95.9 | 83.9 | |
| Yong et al.year=20232024.05 | 92 | 91.4 | 65.8 | 91.6 | 83.1 | |
| ROSVOTvariant=conv2024.05 | 91.6 | 92.6 | 70.9 | 96.8 | 84.1 | |
| ROSVOTablation=w/o wbd2024.05 | 91.3 | 91.8 | 70.2 | 95.5 | 83.8 | |
| MusicYOLO2024.05 | 82.2 | 81.7 | 58.9 | 85.4 | 81.6 | |
| VOCANO2024.05 | 75.8 | 71.2 | 50.2 | 81.4 | 76.6 | |
| TONY2024.05 | 67.5 | 57.8 | 43.9 | 73.8 | 73.9 |