Speech Separation on VoxCeleb2 2-speaker mixture, On-screen speaker (test)
8.46SISDRRAVSS
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| RAVSSON-SCREEN=true, UNSUP.=false2026.02 | 8.46 | 2.42 | 0.65 | 2.02 | 25.1 | 24.4 | |
| SSNAPSON-SCREEN=true, UNSUP.=true2026.02 | 6.24 | 2.17 | 0.64 | 2.64 | 25.5 | 26.1 | |
| SSNAPS w/o L_CT-SSON-SCREEN=true, UNSUP.=true, ablation=without L_CT-SS2026.02 | 6.07 | 2.11 | 0.62 | 2.46 | 27.3 | 26.9 | |
| MIXTUREON-SCREEN=true2026.02 | -2.54 | 1.48 | 0.38 | 1.13 | 93 | 84.7 |