Audio-Visual Speech Recognition on LRS3 Babble noise (Multi-SNR CER Evaluation)
23.47CER (-10 dB)AV-HuBERT Large
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| AV-HuBERT LargeBackbone=Large, Noise Type=Babble2026.04 | 23.47 | 10.57 | 3.6 | 1.7 | 1.21 | |
| VisG AV-HuBERT LargeBackbone=Large, Noise Type=Babble2026.04 | 24.25 | 11.1 | 3.66 | 1.74 | 1.24 | |
| AV-HuBERT BaseBackbone=Base, Noise Type=Babble2026.04 | 31.5 | 17.25 | 7.09 | 3.27 | 2.5 | |
| VisG AV-HuBERT BaseBackbone=Base, Noise Type=Babble2026.04 | 32.43 | 16.15 | 6.66 | 3.5 | 2.49 |