Vocal Effort Classification on AVID vocal effort corpus (10-fold CV)
78.22Mean AccuracyMixUp + GN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MixUp + GNBackbone=WavLM-Base, Augmentation=MixUp, Alpha=0.6, Labeling Strategy=Gaussian-Neighbor soft labels2026.06 | 78.22 | 1.18 | |
| GN (Soft labels)Backbone=WavLM-Base, Labeling Strategy=Gaussian-Neighbor soft labels2026.06 | 77.32 | 1.46 | |
| MixUp + GNBackbone=WavLM-Base, Augmentation=MixUp, Labeling Strategy=Gaussian-Neighbor soft labels2026.06 | 77.27 | 1.47 | |
| CutMix + GNBackbone=WavLM-Base, Augmentation=CutMix, Labeling Strategy=Gaussian-Neighbor soft labels2026.06 | 77.18 | 1.32 | |
| MixUpBackbone=WavLM-Base, Augmentation=MixUp2026.06 | 77 | 1.52 | |
| Label smoothingBackbone=WavLM-Base, Labeling Strategy=Label smoothing2026.06 | 76.95 | 1.45 | |
| CutMixBackbone=WavLM-Base, Augmentation=CutMix2026.06 | 76.91 | 1.48 | |
| MixUp + GNBackbone=WavLM-Base, Augmentation=MixUp, Alpha=0.8, Labeling Strategy=Gaussian-Neighbor soft labels2026.06 | 76.85 | 1.29 | |
| Hard labelsBackbone=WavLM-Base, Labeling Strategy=Hard labels2026.06 | 75.24 | 1.47 |