List your company
ResearchFormerly WizWand

Image Classification on ImageNet-1k (Accuracy & Hardware Efficiency)

79.82AccuracyGe²mS-T Large
63.07667.42371.7776.117Apr 10, 2026Apr 11, 2026Apr 12, 2026Apr 13, 2026Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Ge²mS-T LargeType=Ge²mS-T Training, Architecture=Ge²mS-T Large, Param.(M)=14.48, T=4, Input size=224x2242026.0479.823.152.83
Ge²mS-T BaseType=Ge²mS-T Training, Architecture=Ge²mS-T Base, Param.(M)=9.36, T=4, Input size=224x2242026.0478.192.141.92
SpikingformerType=STBP Training, Architecture=Spikingformer-8-768, Param.(M)=66.34, T=4, Input size=224x2242026.0477.64—16.3
Att-MS ResNetType=STBP Training, Architecture=MS ResNet-104†, Param.(M)=78.37, T=4, Input size=288x2882026.0477.08—7.3
SNN-ViTType=STBP Training, Architecture=SViT-8-384, Param.(M)=30.4, T=4, Input size=224x2242026.0476.87—20.83
SpikingformerType=STBP Training, Architecture=Spikingformer-8-512, Param.(M)=29.68, T=4, Input size=224x2242026.0476.54—8.68
Spike-driven TransformerType=STBP Training, Architecture=SDT-8-768, Param.(M)=66.34, T=4, Input size=224x2242026.0476.32—6.09
MS ResNetType=STBP Training, Architecture=MS ResNet-104†, Param.(M)=77.28, T=5, Input size=288x2882026.0476.0211.3210.19
OursArchitecture=Spikingformer-8-384, Param (M)=16.81, Train Size=224^2, Test Size=224^2, Time Step=42026.0475.53—5.67
NILIF (Spikingformer†)Architecture=Spikingformer-8-384, Param (M)=16.81, Train Size=224^2, Test Size=224^2, Time Step=42026.0475.41—5.58
Ge²mS-T SmallType=Ge²mS-T Training, Architecture=Ge²mS-T Small, Param.(M)=5.35, T=4, Input size=224x2242026.0475.121.291.16
SpikformerType=STBP Training, Architecture=Spikformer-8-768, Param.(M)=66.34, T=4, Input size=224x2242026.0474.8122.0921.48
SNN-ViTType=STBP Training, Architecture=SViT-8-256, Param.(M)=13.7, T=4, Input size=224x2242026.0474.66—14.28
Spike-driven TransformerType=STBP Training, Architecture=SDT-8-512, Param.(M)=29.68, T=4, Input size=224x2242026.0474.57—4.5
SpikingformerType=STBP Training, Architecture=Spikingformer-8-384, Param.(M)=16.81, T=4, Input size=224x2242026.0474.35—5.61
LIF (Spikingformer†)Architecture=Spikingformer-8-384, Param (M)=16.81, Train Size=224^2, Test Size=224^2, Time Step=42026.0474.35—5.61
SpikformerType=STBP Training, Architecture=Spikformer-8-512, Param.(M)=29.68, T=4, Input size=224x2242026.0473.3811.0911.58
Spiking ResNetType=ANN-SNN Conversion, Architecture=ResNet-50, Param.(M)=25.56, T=350, Input size=224x2242026.0472.7578.2970.93
LIF (Spikingformer)Architecture=Spikingformer-8-384, Param (M)=16.81, Train Size=224^2, Test Size=224^2, Time Step=42026.0472.45—4.69
Spike-driven TransformerType=STBP Training, Architecture=SDT-8-384, Param.(M)=16.81, T=4, Input size=224x2242026.0472.28—3.9
SD-TransformerArchitecture=SD-Transformer-8-384, Param (M)=16.81, Train Size=224^2, Test Size=224^2, Time Step=42026.0472.28—3.9
Spiking ResNetType=ANN-SNN Conversion, Architecture=ResNet-34, Param.(M)=21.79, T=350, Input size=224x2242026.0471.6165.2859.3
SpikformerType=STBP Training, Architecture=Spikformer-8-384, Param.(M)=16.81, T=4, Input size=224x2242026.0470.246.827.73
SpikformerArchitecture=Spikformer-8-384, Param (M)=16.81, Train Size=224^2, Test Size=224^2, Time Step=42026.0470.24—12.43
MS ResNetType=STBP Training, Architecture=MS ResNet-34, Param.(M)=21.79, T=6, Input size=224x2242026.0469.424.774.29
MS-ResNetArchitecture=MS-ResNet-34, Param (M)=21.80, Train Size=224^2, Test Size=224^2, Time Step=62026.0469.42——
SEW ResNetType=STBP Training, Architecture=SEW ResNet-152, Param.(M)=60.19, T=4, Input size=224x2242026.0469.2613.7212.89
SEW ResNetType=STBP Training, Architecture=SEW ResNet-101, Param.(M)=44.55, T=4, Input size=224x2242026.0468.769.38.91
SEW ResNetType=STBP Training, Architecture=SEW ResNet-50, Param.(M)=25.56, T=4, Input size=224x2242026.0467.784.834.89
SEW ResNetType=STBP Training, Architecture=SEW ResNet-34, Param.(M)=21.79, T=4, Input size=224x2242026.0467.043.884.04
SEW ResNetArchitecture=SEW-ResNet-34, Param (M)=21.79, Train Size=224^2, Test Size=224^2, Time Step=42026.0467.04——
STBP-tdBNType=STBP Training, Architecture=Spiking ResNet-34, Param.(M)=21.79, T=6, Input size=224x2242026.0463.726.56.39