Automatic Video-Speech Recognition on LRS3 Overlapped speech, SNR=-5dB (test)
0.096WER (%)AVBC
Evaluation Results
| Method | Links | |
|---|---|---|
| AVBCVideo assistance=w/ vid2026.01 | 0.096 | |
| AVBC (w/o enh)Video assistance=w/ vid, speech enhancement=false2026.01 | 0.107 | |
| Unified-AttentionVideo assistance=w/ vid2026.01 | 0.113 | |
| AVBCVideo assistance=w/o vid2026.01 | 0.246 | |
| AVBC (w/o enh)Video assistance=w/o vid, speech enhancement=false2026.01 | 0.259 | |
| Unified-AttentionVideo assistance=w/o vid2026.01 | 0.274 |