SOTA Speech Separation benchmarks and papers with code

Benchmarks

Dataset Name	SOTA Method	Metric
WSJ0-2Mix (test)	TF-Locoformer (L)	SDRi (dB)25.2	141	4d ago
WSJ0-2Mix	SepReformer-L	SI-SNRi (dB)25.1	65	4d ago
WHAM! (test)	MossFormer2	SI-SNRi (dB)18.1	58	4d ago
WHAMR! (test)	TF-GridNet	ΔSI-SNR17.1	57	4d ago
Libri2Mix (test)	SepReformer-M	SI-SNRi (dB)22	45	4d ago
WSJ0-3mix (test)	SepFormer	SI-SNRi19.5	29	4d ago
WHAMR!	TF-Locoformer (M)	SI-SNRi18.5	20	4d ago
WHAM!	SepReformer-L	SI-SNRi (dB)18.4	15	4d ago
VoxCeleb2-2Mix (test)	CTCNet	SDRi13.1	12	4d ago
WSJ0-3mix (clean)	SepFormer	Delta SI-SNR (dB)19.5	12	4d ago
WSJ0-2mix 8 kHz (test)	Sandglasset (MG) + PT	SI-SNRi21	12	4d ago
LRS3-2Mix (test)	CTCNet	SDRi17.5	11	4d ago
Speech (test)	DPRNN	SI-SDRi18.8	11	4d ago
LRS2-2Mix (test)	TDANet	GPU RTF (s) (Forward)0.0118	10	4d ago
WSJ0-4mix (test)	Voice Separation	SI-SNRi12.9	10	4d ago
Libri-10Mix	Generative Upper Bound	SI-SDRi (dB)15	9	4d ago
Libri-5Mix	Generative Upper Bound	SI-SDRi (dB)17.5	9	4d ago
WSJ0 open speaker set (test)		SDR Improvement (dB)8.3	9	4d ago
LibriMix (test)	Sepformer	UTMOS3.54	8	4d ago
VCTK 2 Speech	TF-Locoformer	SI-SDR14.52	7	4d ago
Libri2Mix		SI-SDRi20.4	6	4d ago
Libri2Mix max mode (test)	DPTNET-SRSSN	Delta SI-SNR (dB)18.3	6	4d ago
LibriCSS Utterance-wise Single-channel (test)	WavLM Large	WER (OS)4.2	6	4d ago
LibriCSS Utterance-wise, Seven-channel (test)	BLSTM	Hybrid ASR WER (OS)7	6	4d ago
GRID (test)	UniVoiceLite	SDR1.46	5	4d ago

Showing 25 of 46 rows