Audio-Visual Speech Recognition on LRS2 Babble noise (Multi-SNR CER Evaluation)
30.66CER (-10 dB SNR)VisG AV-HuBERT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VisG AV-HuBERTArchitecture=Large2026.04 | 30.66 | 20.73 | 14.03 | 10.79 | 10.1 | |
| AV-HuBERTArchitecture=Large2026.04 | 32.04 | 21.86 | 13.72 | 11.32 | 10.47 | |
| AV-HuBERTArchitecture=Base2026.04 | 37.92 | 26.04 | 15.74 | 11.59 | 9.62 | |
| VisG AV-HuBERTArchitecture=Base2026.04 | 38.46 | 26.01 | 15.66 | 11.84 | 10.25 |