Timbre-separated transcription on URMP 2mix
69.1F-score (Frame-level)Note-Level Contrastive Clustering
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Note-Level Contrastive ClusteringTraining Data=MusicNet, Embedding Dimension=16, Loss=InfoNCE [8]2025.09 | 69.1 | 68.4 | 84.7 | |
| Note-Level Contrastive ClusteringTraining Data=MusicNet, Coupling Strategy=Shared residual block, Loss=InfoNCE [8]2025.09 | 69 | 68.3 | 84.6 | |
| Note-Level Contrastive ClusteringTraining Data=MusicNet, Loss=InfoNCE [8]2025.09 | 68.9 | 66.8 | 82.6 | |
| Note-Level Contrastive ClusteringTraining Data=MusicNet, Coupling Strategy=Rescale, Loss=InfoNCE [8]2025.09 | 68.4 | 66.8 | 82.7 | |
| TanakaBaseline=Tanaka et al. (2020), Loss=InfoNCE [8]2025.09 | 65.5 | 64.1 | 79.3 | |
| Note-Level Contrastive ClusteringTraining Data=MusicNet, Loss=L_affinity [6]2025.09 | 65.4 | 65.4 | 80.9 | |
| Note-Level Contrastive ClusteringTraining Data=Synthetic, Loss=InfoNCE [8]2025.09 | 49.6 | 53.9 | 66.7 |