List your company
ResearchFormerly WizWand

Keyword Spotting on Google Speech Commands v1 (test)

98AccuracyBC-ResNet-8
1.467226.528651.5976.6514Aug 27, 2018Apr 16, 2019Dec 5, 2019Jul 25, 2020Mar 14, 2021Nov 2, 2021Jun 23, 2022
Updated 8mo ago

Evaluation results

MethodLinks
BC-ResNet-8#Param=321k, #Mult=89.1M2021.0698
BC-ResNet-6#Param=188k, #Mult=53.1M2021.0697.9
BC-ResNet-3#Param=54.2k, #Mult=16.2M2021.0697.6
MatchboxNet-3x2x64#Param=93k, #Mult=11.3M2021.0697.5
BC-ResNet-2#Param=27.3k, #Mult=8.5M2021.0697.3
MHAtt-RNNimplementation=Optimized (SpecAugment + tuning)2020.0597.2
BC-ResNet-1.5#Param=17.2k, #Mult=5.5M2021.0697.2
MatchboxNet-3x1x64#Param=77k, #Mult=9.3M2021.0697.2
MHAtt-RNN#Param=743k, #Mult=22.7M2021.0697.2
TC-ResNetimplementation=Optimized (SpecAugment + tuning)2020.0597.1
DSCNN+strdimplementation=Optimized (SpecAugment + tuning)2020.0597
CRNNimplementation=Optimized (SpecAugment + tuning)2020.0597
Attention RNNTrainable Parameters=202K, Model Variant=V22018.0896.9
LSTMimplementation=Optimized (SpecAugment + tuning)2020.0596.9
EdgeSpeechNet-A# Parameters, K=1072020.0496.8
DenseNet-BC-100# Parameters, K=8002020.0496.77
DS-ResNet18#Param=72 k, #Mult=285 M2021.0696.7
GRUimplementation=Optimized (SpecAugment + tuning)2020.0596.6
TC-ResNetimplementation=Baseline Reference2020.0596.6
TC-ResNet14-1.5#Param=305k, #Mult=6.7M2021.0696.6
TENet12#Param=100k, #Mult=2.9M2021.0696.6
BC-ResNet-1#Param=9.2k, #Mult=3.1M2021.0696.6
SVDFimplementation=Optimized (SpecAugment + tuning)2020.0596.3
S4D-LinParameters=306K, Sampling Rate=16000Hz, Zero-shot=false2022.0696.25
SSAST 250Masked Patches=250, Pretraining=Self-Supervised2021.1096.2
S4D-InvParameters=306K, Sampling Rate=16000Hz, Zero-shot=false2022.0696.18
S4-LegSParameters=307K, Sampling Rate=16000Hz, Zero-shot=false2022.0696.08
SSAST 400Masked Patches=400, Pretraining=Self-Supervised2021.1096
DS-ResNet14#Param=15.2k, #Mult=15.7M2021.0695.9
S4D-LegSParameters=306K, Sampling Rate=16000Hz, Zero-shot=false2022.0695.83
res15Trainable Parameters=238K2018.0895.8
ResNet-15#Param=238k, #Mult=894M2021.0695.8
Attention RNNTrainable Parameters=202K2018.0895.6
Att-RNNimplementation=Baseline Reference2020.0595.6
Att-RNN#Param=202k, #Mult=22.3M2021.0695.6
ConvNetParameters=26.2M, Sampling Rate=16000Hz, Zero-shot=false2022.0695.51
AST-IM + KDPretraining=ImageNet + Knowledge Distillation2021.1095.5
DS-CNNTrainable Parameters=498K2018.0895.4
CNN+strdimplementation=Optimized (SpecAugment + tuning)2020.0595.4
DSCNN+strdimplementation=Baseline Reference2020.0595.4
S4-(LegS/FouT)Parameters=307K, Sampling Rate=16000Hz, Zero-shot=false2022.0695.32
S4-FouTParameters=307K, Sampling Rate=16000Hz, Zero-shot=false2022.0695.27
res26Trainable Parameters=438K2018.0895.2
CRNNimplementation=Baseline Reference2020.0595
LSTMimplementation=Baseline Reference2020.0594.8
GRUimplementation=Baseline Reference2020.0594.7
res8Trainable Parameters=110K2018.0894.1
CNN+strdimplementation=Baseline Reference2020.0592.7
S4D-InvParameters=306K, Sampling Rate=8000Hz, Zero-shot=true2022.0691.8
AST-AudioSetPretraining=Supervised AudioSet2021.1091.6
S4-FouTParameters=307K, Sampling Rate=8000Hz, Zero-shot=true2022.0691.59
S4D-LinParameters=306K, Sampling Rate=8000Hz, Zero-shot=true2022.0691.58
S4-LegSParameters=307K, Sampling Rate=8000Hz, Zero-shot=true2022.0691.32
DNNimplementation=Optimized (SpecAugment + tuning)2020.0591.2
S4D-LegSParameters=306K, Sampling Rate=8000Hz, Zero-shot=true2022.0691.08
S4-(LegS/FouT)Parameters=307K, Sampling Rate=8000Hz, Zero-shot=true2022.0690.72
ConvNet on raw WAVTrainable Parameters=700K2018.0889.4
AST-ScratchTraining=From scratch2021.1087.2
DNNimplementation=Baseline Reference2020.0586.7
XResNet-50Parameters=904K, Sampling Rate=16000Hz, Zero-shot=false2022.0683.01
ResNet-18Parameters=216K, Sampling Rate=16000Hz, Zero-shot=false2022.0677.86
TRILLEvaluation Protocol=Linear Classifier2020.1074
COLAEvaluation Protocol=Linear Classifier2020.1071.7
InceptionNetParameters=481K, Sampling Rate=16000Hz, Zero-shot=false2022.0661.24
ResNet-18Parameters=216K, Sampling Rate=8000Hz, Zero-shot=true2022.068.74
XResNet-50Parameters=904K, Sampling Rate=8000Hz, Zero-shot=true2022.067.72
ConvNetParameters=26.2M, Sampling Rate=8000Hz, Zero-shot=true2022.067.26
InceptionNetParameters=481K, Sampling Rate=8000Hz, Zero-shot=true2022.065.18