Audio-Visual Speech Separation on VoxCeleb2 (off-screen speakers)
7.12SISDRRAVSS
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RAVSSON-SCREEN=false, UNSUPERVISED=false2026.02 | 7.12 | 2.23 | 0.62 | 1.85 | 28.9 | 27 | |
| SSNAPSON-SCREEN=false, UNSUPERVISED=true2026.02 | 4.28 | 1.81 | 0.56 | 2.71 | 55.1 | 55.1 | |
| SSNAPS w/o L_CT-SSON-SCREEN=false, UNSUPERVISED=true2026.02 | 3.71 | 1.71 | 0.55 | 2.59 | 60.5 | 58.7 | |
| MIXTUREON-SCREEN=false, UNSUPERVISED=N/A2026.02 | -3.08 | 1.44 | 0.33 | 0.87 | 208 | 205.8 |