Automatic Speech Recognition on LRS2-2Mix (test)
17.74WERGround-truth
Evaluation Results
| Method | Links | |
|---|---|---|
| Ground-truth2022.12 | 17.74 | |
| Ground-truth2023.09 | 17.74 | |
| CTCNetASR API=Google Cloud Speech-to-Text2022.12 | 24.82 | |
| CTCNet2023.09 | 24.82 | |
| RTFS-Net-12variant=12 blocks2023.09 | 24.93 | |
| RTFS-Net-8variant=8 blocks2023.09 | 27.42 | |
| RTFS-Net-4variant=4 blocks2023.09 | 29.66 | |
| AV-ConvTasNetASR API=Google Cloud Speech-to-Text2022.12 | 31.43 | |
| AV-ConvTasNet2023.09 | 31.43 | |
| AVLIT-8ASR API=Google Cloud Speech-to-Text2022.12 | 31.85 | |
| AVLIT-8variant=8 blocks2023.09 | 31.85 | |
| CaffNet-CPhase Information Usage=GT phase + estimated amplitude, ASR API=Google Cloud Speech-to-Text2022.12 | 32.96 | |
| CaffNet-C2023.09 | 32.96 | |
| VisualVoiceASR API=Google Cloud Speech-to-Text2022.12 | 34.45 | |
| Visualvoice2023.09 | 34.45 | |
| CaffNet-CPhase Information Usage=Standard, ASR API=Google Cloud Speech-to-Text2022.12 | 35.08 | |
| Mixture2022.12 | 84.91 | |
| Mixture2023.09 | 84.91 |