TitaNet: Neural Model for speaker representation with 1D Depth-wise separable convolutions and global context

Open paper
Benchmark Model Rank Results
speaker-diarization-on-ami-lapelTitaNet-M (NME-SC)#1DER(%): 1.99
speaker-diarization-on-ami-lapelTitaNet-S (NME-SC)#2DER(%): 2.00
speaker-diarization-on-ami-lapelTitaNet-L (NME-SC)#3DER(%): 2.03
speaker-diarization-on-ami-lapelECAPA (SC)#4DER(%): 2.36
speaker-diarization-on-ami-mixheadsetTitaNet-L (NME-SC)#1DER(%): 1.73
speaker-diarization-on-ami-mixheadsetECAPA (SC)#2DER(%): 1.78
speaker-diarization-on-ami-mixheadsetTitaNet-M (NME-SC)#3DER(%): 1.79
speaker-diarization-on-ami-mixheadsetTitaNet-S (NME-SC)#4DER(%): 2.22
speaker-diarization-on-ch109TitaNet-S (NME-SC)#1DER(%): 1.11
speaker-diarization-on-ch109TitaNet-M (NME-SC)#2DER(%): 1.13
speaker-diarization-on-ch109TitaNet-L (NME-SC)#3DER(%): 1.19
speaker-diarization-on-ch109x-vector (PLDA + AHC)#4DER(%): 9.72
speaker-diarization-on-nist-sre-2000x-vector (MCGAN)#1DER(%): 5.73
speaker-diarization-on-nist-sre-2000TitaNet-S (NME-SC)#2DER(%): 6.37
speaker-diarization-on-nist-sre-2000TitaNet-M (NME-SC)#3DER(%): 6.47
speaker-diarization-on-nist-sre-2000TitaNet-L (NME-SC)#4DER(%): 6.73
speaker-diarization-on-nist-sre-2000x-vector (PLDA + AHC)#5DER(%): 8.39
speaker-verification-on-voxceleb1-o-cleanTitaNet-L (extra data)#5EER: 0.68minDCF (P_target=0.01): 0.087
speaker-verification-on-voxcelebTitanNet -L#13EER: 0.68
speaker-verification-on-voxcelebTitanNet -M#15EER: 0.81
speaker-verification-on-voxcelebTitanNet -S#19EER: 1.15

Rank counts only results with a code link.