List your company
ResearchFormerly WizWand

Anti-spoofing on ASVspoof LA 2019 (test)

1.54EERFastAudio-Tri
0.79725.811110.82515.8389Sep 5, 2021Sep 6, 2021
Updated 8mo ago

Evaluation results

MethodLinks
FastAudio-TriBackend=ECAPA-TDNN, #Params=80, Constraint=Shape+Clamp, MACS=0.00GMac, Train Time/Epoch=13:02 min2021.091.540.0451
FastAudio-GaussBackend=ECAPA-TDNN, #Params=80, Constraint=Shape+Clamp, MACS=0, Train Time/Epoch=12:51 min2021.091.630.0471
FastAudio-GaussBackend=X-vector, #Params=80, Constraint=Shape+Clamp, MACS=0, Train Time/Epoch=12:51 min2021.091.670.0516
FastAudio-SortBackend=X-vector, #Params=80, Constraint=Shape+Clamp+Order, MACS=0, Train Time/Epoch=12:59 min2021.091.690.0524
CQTBackend=ECAPA-TDNN, #Params=0, Constraint=Fixed, MACS=0, Train Time/Epoch=10:58 min2021.091.730.0508
FastAudio-TriBackend=X-vector, #Params=80, Constraint=Shape+Clamp, MACS=0.00GMac, Train Time/Epoch=13:02 min2021.091.730.0491
TD-filterbanksBackend=ECAPA-TDNN, #Params=31k, Constraint=Shape+Clamp, MACS=1.32GMac, Train Time/Epoch=22.48 min2021.091.830.0528
FastAudio-SortBackend=ECAPA-TDNN, #Params=80, Constraint=Shape+Clamp+Order, MACS=0, Train Time/Epoch=12:59 min2021.091.890.052
LFCC+SE-ResABNet+CombLossInput feature=LFCC, Classifier=SE-ResABNet, Loss function=CombLoss2021.091.890.0507
FbanksBackend=ECAPA-TDNN, #Params=0, Constraint=Fixed, MACS=0, Train Time/Epoch=10:53 min2021.092.110.0643
LFCC+ResNet18+OCSInput feature=LFCC, Classifier=ResNet18, Loss function=OCS2021.092.190.059
FbanksBackend=X-vector, #Params=0, Constraint=Fixed, MACS=0, Train Time/Epoch=10:53 min2021.092.390.0688
LEAFBackend=ECAPA-TDNN, #Params=282, Constraint=Shape+Clamp, MACS=0.01GMac, Train Time/Epoch=34.45 min2021.092.490.0645
CQT+SE-Res2Net50+CEInput feature=CQT, Classifier=SE-Res2Net50, Loss function=CE2021.092.50.0743
LFCC+SE-Res2Net50+CEInput feature=LFCC, Classifier=SE-Res2Net50, Loss function=CE2021.092.870.0786
TD-filterbanksBackend=X-vector, #Params=31k, Constraint=Shape+Clamp, MACS=1.32GMac, Train Time/Epoch=22.48 min2021.093.180.0843
LEAFBackend=X-vector, #Params=282, Constraint=Shape+Clamp, MACS=0.01GMac, Train Time/Epoch=34.45 min2021.093.280.0732
CQTBackend=X-vector, #Params=0, Constraint=Fixed, MACS=0, Train Time/Epoch=10:58 min2021.093.40.0951
nnAudioBackend=ECAPA-TDNN, #Params=8.04k, Constraint=No, MACS=0, Train Time/Epoch=13:00 min2021.093.630.0893
Spect+LCGRNN+GKDE-softmaxInput feature=Spect, Classifier=LCGRNN, Loss function=GKDE-softmax2021.093.770.0842
FG-CQT+LCNN+CEInput feature=FG-CQT, Classifier=LCNN, Loss function=CE2021.094.070.102
Spect+LCNN+A-softmaxInput feature=Spect, Classifier=LCNN, Loss function=A-softmax2021.094.530.1028
LFCC+LCNN+A-softmaxInput feature=LFCC, Classifier=LCNN, Loss function=A-softmax2021.095.060.1
SincNetBackend=Baseline, Name=RawNet2, Constraint=Fixed2021.095.130.1175
nnAudioBackend=X-vector, #Params=8.04k, Constraint=No, MACS=0, Train Time/Epoch=13:00 min2021.095.560.1471
LFCC+GMM+EMInput feature=LFCC, Classifier=GMM, Loss function=EM2021.098.090.2116
Spect+SE-Res2Net50+CEInput feature=Spect, Classifier=SE-Res2Net50, Loss function=CE2021.098.730.2237
MFCC+ResNet+CEInput feature=MFCC, Classifier=ResNet, Loss function=CE2021.099.330.2042
CQCC+GMM+EMInput feature=CQCC, Classifier=GMM, Loss function=EM2021.099.570.2366
Spect+ResNet+CE [29]Input feature=Spect, Classifier=ResNet, Loss function=CE2021.099.680.2741
Spect+ResNet+CE [20]Input feature=Spect, Classifier=ResNet, Loss function=CE2021.0911.750.216
Raw signal+SincNet+CEInput feature=Raw signal, Classifier=SincNet, Loss function=CE2021.0920.110.3563