Language Recognition on TidyVoice
87.3Top-1 AccuracyCLIP ViT-B/32
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CLIP ViT-B/32Backbone=ViT-B/322026.06 | 87.3 | — | |
| DinoV3 ViT-SBackbone=ViT-S2026.06 | 83.1 | — | |
| PlanFormerProtocol=Linear probing (5-fold CV), Parameters=1.99M, FLOPs=93.6MFLOPs, Classes=29, Type=Semantic2026.06 | 69.8 | 91.2 | |
| Expert features (MFCC) + linear SVMType=Expert Features2026.06 | 69.7 | — | |
| Random Init.Classes=29, Type=Semantic2026.06 | 49.9 | — |