Visual Speech Recognition on WildVSR
38.6WERAuto-AVSR
Evaluation Results
| Method | Links | |
|---|---|---|
| Auto-AVSRLabelled hours=3448, Shared params=false2024.11 | 38.6 | |
| USRLabelled hours=433, Unlabelled hours=1326, Shared params=true2024.11 | 46.4 | |
| RAVEn w/ self-trainingLabelled hours=433, Unlabelled hours=1326, Shared params=false2024.11 | 46.7 | |
| AV-HuBERT w/ self-trainingLabelled hours=433, Unlabelled hours=1326, Shared params=false2024.11 | 48.7 | |
| Auto-AVSRLabelled hours=1759, Shared params=false2024.11 | 49.3 | |
| AV-HuBERTLabelled hours=433, Unlabelled hours=1326, Shared params=false2024.11 | 51.7 | |
| RAVEnLabelled hours=433, Unlabelled hours=1326, Shared params=false2024.11 | 52.2 | |
| CM-seq2seqLabelled hours=1459, Shared params=false2024.11 | 58.4 | |
| Auto-AVSRLabelled hours=661, Shared params=false2024.11 | 62.3 | |
| VTPLabelled hours=2676, Shared params=false2024.11 | 68.7 | |
| USR 2.0Decoding=greedy2026.02 | 73.7 | |
| VTPLabelled hours=698, Shared params=false2024.11 | 75.6 | |
| USRDecoding=greedy2026.02 | 80 | |
| BRAVEnDecoding=greedy2026.02 | 81.2 | |
| AV-HuBERTDecoding=greedy2026.02 | 82.4 |