Visual Speech Recognition on CMLR (test)
52.3Inference Latency (ms)Lipnet
Evaluation results
| Method | Links | |
|---|---|---|
| LipnetStructure=single-stage, LSTM, #Params=18.16 M2026.03 | 52.3 | |
| Cascade-Free Mandarin VSR (F)Structure=cascade-free, Conformer, #Params=86.79 M2026.03 | 79.1 | |
| Cascade-Free Mandarin VSR (F + V)Structure=cascade-free, Conformer, #Params=108.18 M2026.03 | 84.7 | |
| Cascade-Free Mandarin VSR (F + P)Structure=cascade-free, Conformer, #Params=108.20 M2026.03 | 87.6 | |
| CTCH-LipNetStructure=dual-stage, Transformer, #Params=40.14 M2026.03 | 92.7 | |
| Cascade-Free Mandarin VSR (F + P + V)Structure=cascade-free, Conformer, #Params=129.58 M2026.03 | 95.2 | |
| CSSMCMStructure=three-stage, LSTM, #Params=38.45 M2026.03 | 137.2 | |
| CT-MIR-NetStructure=dual-stage, Transformer, #Params=122.57 M2026.03 | 160.9 |