Speaker Diarization on Bengali-Loop
23.92DERBangla-WhisperDiar
Evaluation results
| Method | Links | |
|---|---|---|
| Bangla-WhisperDiarMode=fine-tuned2026.05 | 23.92 | |
| Proposed Dual-VADAcoustic Approach=Finetuned + Exclusive, Source=Proposed2026.03 | 24 | |
| Pyannote.audio (Base)Acoustic Approach=EEND Hybrid, Source=Baseline2026.03 | 40.08 | |
| pyannote.audioState=pretrained2026.05 | 40.08 | |
| Silero VAD + ECAPAAcoustic Approach=Modular Pipeline, Source=Baseline2026.03 | 61.5 | |
| Silero VAD + ECAPA + Clustering2026.05 | 61.5 | |
| WebRTC VAD + ECAPAAcoustic Approach=Modular Pipeline, Source=Baseline2026.03 | 73.71 | |
| WebRTC VAD + ECAPA + Clustering2026.05 | 73.71 |