Image Classification on ImageNet 1K (val) (Top-1 Accuracy and Power)
82.96Top-1 AccuracyMSViT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MSViTType=SNN, Architecture=MSViT-10-512, Param (M)=30.23, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 82.96 | 24.74 | |
| DeiTType=ANN, Architecture=DeiT-B, Param (M)=86.59, Time Step=1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 81.8 | 80.5 | |
| SAFformerType=SNN, Architecture=SAFformer-10-512, Param (M)=26.58, Time Step=1×4, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 80.44 | 5.88 | |
| MSViTType=SNN, Architecture=MSViT-10-384, Param (M)=17.69, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 80.09 | 16.65 | |
| Max-FormerType=SNN, Architecture=Max-10-512, Param (M)=28.65, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 79.86 | 7.49 | |
| DeiTType=ANN, Architecture=DeiT-S, Param (M)=22.00, Time Step=1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 79.8 | 21.2 | |
| E-SpikeFormerType=SNN, Architecture=E-SpikeFormer-M, Param (M)=19.00, Time Step=1×4, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 79.8 | 5.9 | |
| SAFformerType=SNN, Architecture=SAFformer-10-384, Param (M)=15.12, Time Step=1×4, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 79.19 | 3.75 | |
| QSD-TransformerType=SNN, Architecture=S-Transformer v2-M†, Param (M)=3.90, Time Step=1×4, Input Size=224 x 224, Knowledge Distillation=true2026.05 | 78.9 | 5.7 | |
| STAType=A2S, Architecture=ViT-B/32, Param (M)=86.00, Time Step=32, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 78.72 | — | |
| MSTType=A2S, Architecture=Swin Transformer-T, Param (M)=28.50, Time Step=512, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 78.51 | — | |
| E-SpikeFormerType=SNN, Architecture=E-SpikeFormer-S, Param (M)=10.00, Time Step=1×4, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 78.5 | 3 | |
| Max-FormerType=SNN, Architecture=Max-10-384, Param (M)=16.23, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 77.82 | 4.89 | |
| QSD-TransformerType=SNN, Architecture=S-Transformer v2-T†, Param (M)=1.80, Time Step=1×4, Input Size=224 x 224, Knowledge Distillation=true2026.05 | 77.5 | 2.5 | |
| S-Transformer v2Type=SNN, Architecture=S-Transformer v2-8-512, Param (M)=31.30, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 77.2 | 32.8 | |
| S-Transformer v2Type=SNN, Architecture=S-Transformer v2-8-384, Param (M)=15.10, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 74.1 | 16.7 | |
| SpikformerType=SNN, Architecture=Spikformer-8-512, Param (M)=29.68, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 73.38 | 11.58 | |
| SpikformerType=SNN, Architecture=Spikformer-8-384, Param (M)=16.81, Time Step=4×1, Input Size=224 x 224, Knowledge Distillation=false2026.05 | 70.24 | 7.73 |