Audio-Visual Speech Recognition on LRS3 Random noise
6.49CER (-10 dB)VisG AV-HuBERT Large
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| VisG AV-HuBERT LargeBackbone=Large, Noise Type=Noise2026.04 | 6.49 | 3.12 | 1.61 | 1.2 | 0.89 | |
| AV-HuBERT LargeBackbone=Large, Noise Type=Noise2026.04 | 7.54 | 3.49 | 1.75 | 1.16 | 0.98 | |
| VisG AV-HuBERT BaseBackbone=Base, Noise Type=Noise2026.04 | 10.32 | 5.85 | 3.14 | 2.6 | 2.32 | |
| AV-HuBERT BaseBackbone=Base, Noise Type=Noise2026.04 | 10.65 | 5.42 | 3.51 | 2.78 | 2.2 |