Conv-Tasnet (Libri1Mix speech enhancement pre-trained)

Reported on 2 benchmarks across 1 task · 1 paper · 2 SOTA

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

Speech2 results

Speech SeparationonLibri2Mix
SDRi· uses extra data· 2020-10-29
14.6
best: 22.2 (TF-Locoformer (M))
SOTA
Stabilizing Label Assignment for Speech Separation by Self-supervised Pre-training arXiv:2010.15366
Speech SeparationonLibri2Mix
SI-SDRi· uses extra data· 2020-10-29
14.1
best: 22.2 (MossFormer2 (w speed perturb))
SOTA
Stabilizing Label Assignment for Speech Separation by Self-supervised Pre-training arXiv:2010.15366