Audio-Visual Speech Recognition on LRS2 Speech noise (CER by SNR)
17.79CER (-10 dB)VisG AV-HuBERT
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VisG AV-HuBERTArchitecture=Large2026.04 | 17.79 | 12.97 | 11.26 | 10 | 9.88 | |
| AV-HuBERTArchitecture=Large2026.04 | 20.51 | 14.33 | 11.31 | 10.48 | 10.11 | |
| VisG AV-HuBERTArchitecture=Base2026.04 | 21.92 | 16.21 | 12.74 | 11.17 | 10.26 | |
| AV-HuBERTArchitecture=Base2026.04 | 28.71 | 18.09 | 13.25 | 10.62 | 9.75 |