Speaker Verification on VoxCeleb1 hard (test)
1.857EERRecXi
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| RecXiAugmentation=true, Parameters (M)=7.06, system_id=#92023.10 | 1.857 | — | — | 0.179 | |
| Xi + tResNet (baseline)Augmentation=true, Parameters (M)=6.54, system_id=#82023.10 | 1.942 | — | — | 0.186 | |
| MFA-TDNN (standard)Augmentation=true, Parameters (M)=7.322023.10 | 2.049 | — | — | 0.19 | |
| RecXiAugmentation=false, Parameters (M)=7.06, system_id=#52023.10 | 2.097 | — | — | 0.196 | |
| Xi + tResNet (baseline)Augmentation=false, Parameters (M)=6.54, system_id=#42023.10 | 2.169 | — | — | 0.202 | |
| MFA-TDNN (lite)Augmentation=true, Parameters (M)=5.932023.10 | 2.174 | — | — | 0.199 | |
| UP-LS UP-PLDA2023.10 | 2.18 | — | — | 0.224 | |
| H/ASP (AP+softmax)Augmentation=true, Parameters (M)=8.02023.10 | 2.21 | — | — | — | |
| ResNet34 (re-implemented)Augmentation=true, Parameters (M)=6.63, system_id=#72023.10 | 2.221 | — | — | 0.208 | |
| ECAPA-TDNN (re-implemented)Augmentation=true, Parameters (M)=6.19, system_id=#62023.10 | 2.249 | — | — | 0.223 | |
| TSP + tResNet (baseline)Augmentation=false, Parameters (M)=6.21, system_id=#32023.10 | 2.257 | — | — | 0.204 | |
| ECAPA-TDNN (MBFA-MW)Augmentation=true2023.10 | 2.31 | — | — | 0.222 | |
| ECAPA-TDNNAugmentation=true, Parameters (M)=6.22023.10 | 2.32 | — | — | 0.218 | |
| H/ASP (AAM-softmax)Augmentation=true, Parameters (M)=8.02023.10 | 2.49 | — | — | — | |
| SI-Net50Augmentation=false2023.10 | 2.5 | — | — | — | |
| ResNet34 (re-implemented)Augmentation=false, Parameters (M)=6.63, system_id=#22023.10 | 2.5 | — | — | 0.224 | |
| RSKNet-MTSPAugmentation=false, Parameters (M)=13.92023.10 | 2.52 | — | — | 0.237 | |
| ECAPA-TDNN (re-implemented)Augmentation=false, Parameters (M)=6.19, system_id=#12023.10 | 2.674 | — | — | 0.235 | |
| Res2Net-26w8sAugmentation=true, Parameters (M)=9.32023.10 | 2.72 | — | — | 0.272 | |
| ResNet34-SKDFEAugmentation=false, Parameters (M)=5.982023.10 | 2.76 | — | — | 0.278 | |
| H/ASP (AP+softmax)Augmentation=false, Parameters (M)=8.02023.10 | 2.77 | — | — | — | |
| H/ASP (AAM-softmax)Augmentation=false, Parameters (M)=8.02023.10 | 2.78 | — | — | — | |
| AAM-softmax (x-vector)Model=TDNN (x-vector), Loss=AAM-softmax, Training set=VoxCeleb22019.06 | 4.732 | — | — | — | |
| Xie et al.Model=Thin ResNet-34, Loss=Softmax, Training set=VoxCeleb22019.06 | 5.06 | — | — | — | |
| Chung et al.Model=ResNet-50, Loss=Softmax + Contrastive, Training set=VoxCeleb22019.06 | 7.33 | — | — | — | |
| ECAPA-TDNNinput=40-dimensional filterbanks, loss=AAM softmax2021.09 | — | 3.1 | 0.05 | — | |
| w2v2-aampooling=mean&std, loss=AAM softmax, feature encoder=frozen2021.09 | — | 4.33 | 0.08 | — | |
| w2v2-bcescoring=computes scores directly, feature encoder=frozen2021.09 | — | 11.34 | 0.83 | — | |
| w2v2-cepooling=mean&std, feature encoder=frozen2021.09 | — | 4.91 | 0.13 | — | |
| x-vectorinput=40-dimensional filterbanks2021.09 | — | 8.75 | 0.05 | — |