List your company

Audio-Visual Target Speaker Extraction on LRS2 2-mix (test)

3.16DNSMOSLibri2Mix GT
1.99522.29762.62.9024Mar 20, 2026

Evaluation results

MethodLinks
Libri2Mix GT2026.033.16—3.93
TF-GridNetBackbone=TF-GridNet, # params=14.4M2026.032.814.814.32
TF-GridNet + LSM (Ours)Backbone=TF-GridNet, AO backbone frozen=true, # params=2.0M2026.032.7914.794.29
MossFormer2Backbone=MossFormer2, # params=55.7M2026.032.7912.683.47
MossFormer2 + LSM (Ours)Backbone=MossFormer2, AO backbone frozen=true, # params=1.6M2026.032.7912.653.47
AV-MossFormer2# params=57.3M2026.032.6415.523.06
TF-GridNet + ResidualBackbone=TF-GridNet, AO backbone frozen=true, # params=2.6M2026.032.5613.093.2
MossFormer2 + ResidualBackbone=MossFormer2, AO backbone frozen=false, # params=57.3M2026.032.5115.543.02
AV-TFGridNet# params=9.6M2026.032.5115.13.53
MossFormer2 + ResidualBackbone=MossFormer2, AO backbone frozen=true, # params=1.6M2026.032.4414.022.95
AV-ConvTasNet# params=10.3M2026.032.4411.512.49
AV-DPRNN# params=4.1M2026.032.411.972.58
LRS2-2mix GT2026.032.38—3.19
TF-GridNet + ResidualBackbone=TF-GridNet, AO backbone frozen=false, # params=17.0M2026.032.3613.383.17
Conv-TasNetBackbone=Conv-TasNet, # params=5.1M2026.032.357.122.31
DPRNNBackbone=DPRNN, # params=2.6M2026.032.357.772.37
DPRNN + LSM (Ours)Backbone=DPRNN, AO backbone frozen=true, # params=2.0M2026.032.357.742.36
Conv-TasNet + ResidualBackbone=Conv-TasNet, AO backbone frozen=false, # params=6.6M2026.032.3111.722.57
Conv-TasNet + LSM (Ours)Backbone=Conv-TasNet, AO backbone frozen=true, # params=1.5M2026.032.286.822.22
DPRNN + ResidualBackbone=DPRNN, AO backbone frozen=false, # params=5.2M2026.032.2811.662.43
Conv-TasNet + ResidualBackbone=Conv-TasNet, AO backbone frozen=true, # params=1.5M2026.032.139.372.27
DPRNN + ResidualBackbone=DPRNN, AO backbone frozen=true, # params=2.5M2026.032.047.91.71