Automatic Speech Recognition on Bangla Clean Speech v20 SLR53 Bhasha-bichitra (test)
12.3WERBanglaRobustNet
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| BanglaRobustNetmode=fine-tuned, RTF=0.16, optimizer=AdamW2026.01 | 12.3 | 5.7 | 8.9 | 76.4 | |
| Wav2Vec-BERTnumber of parameters=580M, mode=fine-tuned, optimizer=AdamW2026.01 | 14.42 | 2.67 | 9.37 | 73.5 | |
| Whisper-Large-v2number of parameters=1.55B, mode=fine-tuned, optimizer=AdamW2026.01 | 28.86 | 7.47 | 18.76 | 64 | |
| Whisper-Smallnumber of parameters=242M, mode=fine-tuned, optimizer=AdamW2026.01 | 32.17 | 18.17 | 20.91 | 62 |