List your company
ResearchFormerly WizWand

35-way Speech Classification on Speech Commands 8kHz (test)

95.05AccuracyMRConv, Fourier
1.585225.850150.11574.3799Aug 9, 2022Dec 10, 2022Apr 12, 2023Aug 14, 2023Dec 15, 2023Apr 16, 2024Aug 18, 2024
Updated 8mo ago

Evaluation results

MethodLinks
MRConv, FourierSampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.0895.05
MRConv, FourierArchitecture Class=Convolutional Models, Parameterization=Fourier2024.0895.05
S5Parameters=280K, Input length=8000, Zero-shot=true2022.0894.53
S5Sampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.0894.53
S5Architecture Class=State Space Models2024.0894.53
SGConvSampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.0894.29
SGConvArchitecture Class=Convolutional Models2024.0894.29
S4-LegSParameters=307K, Input length=8000, Zero-shot=true2022.0891.32
S4-LegSSampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.0891.32
S4-LegSArchitecture Class=State Space Models2024.0891.32
S4D-LegSParameters=306K, Input length=8000, Zero-shot=true2022.0891.08
S4D-LegSSampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.0891.08
S4D-LegSArchitecture Class=State Space Models2024.0891.08
Liquid-S4Parameters=224K, Input length=8000, Zero-shot=true2022.0890
Liquid-S4Sampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.0890
Liquid-S4Architecture Class=State Space Models2024.0890
ResNet-1Parameters=216K, Input length=8000, Zero-shot=true2022.088.74
ResNet-18Sampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.088.74
ResNet-1Architecture Class=CNNs2024.088.74
XResNet-50Parameters=904K, Input length=8000, Zero-shot=true2022.087.72
XResNet-50Sampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.087.72
XResNet-50Architecture Class=CNNs2024.087.72
ConvNetParameters=26.2M, Input length=8000, Zero-shot=true2022.087.26
ConvNetSampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.087.26
ConvNetArchitecture Class=CNNs2024.087.26
InceptionNetParameters=481K, Input length=8000, Zero-shot=true2022.085.18
InceptionNetSampling rate=8 kHz, Input length=8,000, Zero-shot evaluation=true2024.085.18
InceptionNetArchitecture Class=CNNs2024.085.18