Separate And Diffuse

Reported on 6 benchmarks across 1 task · 1 paper · 6 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Speech6 results

Speech SeparationonWSJ0-2mix
SI-SDRi· 2023-01-25
23.9
best: 25.1 (TF-Locoformer (L) + DM)
SOTA
Separate And Diffuse: Using a Pretrained Diffusion Model for Improving Source Separation arXiv:2301.10752
Speech SeparationonLibri20Mix
SI-SDRi· 2023-01-25
5.2
SOTA
Separate And Diffuse: Using a Pretrained Diffusion Model for Improving Source Separation arXiv:2301.10752
Speech SeparationonWSJ0-3mix
SI-SDRi· 2023-01-25
20.9
best: 23.7 (SepTDA)
SOTA
Separate And Diffuse: Using a Pretrained Diffusion Model for Improving Source Separation arXiv:2301.10752
Speech SeparationonLibri5Mix
SI-SDRi· 2023-01-25
14.2
SOTA
Separate And Diffuse: Using a Pretrained Diffusion Model for Improving Source Separation arXiv:2301.10752
Speech SeparationonLibri2Mix
SI-SDRi· 2023-01-25
21.5
best: 22.2 (MossFormer2 (w speed perturb))
SOTA
Separate And Diffuse: Using a Pretrained Diffusion Model for Improving Source Separation arXiv:2301.10752
Speech SeparationonLibri10Mix
SI-SDRi· 2023-01-25
9
SOTA
Separate And Diffuse: Using a Pretrained Diffusion Model for Improving Source Separation arXiv:2301.10752