Cinematic Audio Source Separation on DnR (test)
16.3Speech SISDRSFC-Locoformer (Medium)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SFC-Locoformer (Medium)Encoder/Decoder=SFC-CA (64), Params.=16.0 M, Dynamic mixing=true, Musical split=true2026.02 | 16.3 | 16.4 | 9.7 | 9.3 | 10.6 | 9.9 | 12.2 | 11.8 | |
| SFC-Locoformer (Medium)Encoder/Decoder=SFC-Mamba (64), Params.=15.2 M, Dynamic mixing=true, Musical split=true2026.02 | 16.2 | 16.2 | 9.6 | 9.1 | 10.5 | 9.8 | 12.1 | 11.7 | |
| BS-Locoformer (Medium)Encoder/Decoder=Band-split (64) [15], Params.=55.5 M, Dynamic mixing=true, Musical split=true2026.02 | 16.1 | 16.2 | 9.4 | 8.9 | 10.3 | 9.7 | 11.9 | 11.6 | |
| SFC-Locoformer (Medium)Encoder/Decoder=SFC-CA (64), Params.=16.0 M, Dynamic mixing=false, Musical split=true2026.02 | 16.1 | 16.2 | 9.2 | 8.7 | 10.1 | 9.4 | 11.8 | 11.4 | |
| SFC-Locoformer (Small)Encoder/Decoder=SFC-CA (64), Params.=5.8 M, Dynamic mixing=true, Musical split=true2026.02 | 15.9 | 15.9 | 9.3 | 8.8 | 10.2 | 9.4 | 11.8 | 11.4 | |
| BandItEncoder/Decoder=Band-split (64) [15], Params.=37.0 M, Dynamic mixing=false, Musical split=true2026.02 | 15.6 | 15.7 | 8.7 | 8 | 9.8 | 8.2 | 11.4 | 10.9 | |
| SFC-Locoformer (Small)Encoder/Decoder=SFC-Mamba (64), Params.=5.1 M, Dynamic mixing=true, Musical split=true2026.02 | 15.6 | 15.7 | 9 | 8.5 | 10 | 9.3 | 11.6 | 11.1 | |
| BS-Locoformer (Small)Encoder/Decoder=Band-split (64) [15], Params.=34.7 M, Dynamic mixing=true, Musical split=true2026.02 | 15.5 | 15.6 | 8.8 | 8.3 | 9.8 | 9.1 | 11.4 | 11 | |
| Hybrid Demucs (v3)Dynamic mixing=false, Params.=83.6 M2026.02 | 13.4 | 13.6 | 6 | 4.7 | 7.2 | 6.1 | 8.9 | 8.1 | |
| MRX-CDynamic mixing=false2026.02 | 12.6 | — | — | 4.6 | — | 6.1 | — | 7.8 |