Gesture Recognition on DVS Gesture (test)
99.3AccuracySpike-driven Transformer
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Spike-driven TransformerType=Direct Training, T=162024.03 | 99.3 | — | — | — | — | — | |
| CogniSNNNetwork=ER-RGA-7, Param(M)=0.13, T=162025.12 | 98.61 | — | — | — | — | — | |
| CMLNetwork=Spikformer-4-384, Param(M)=2.57, T=162025.12 | 98.6 | — | — | — | — | — | |
| STSANetwork=STS-Transformer-1-256, T=162025.12 | 98.38 | — | — | — | — | — | |
| SpikingformerType=Direct Training, T=162024.03 | 98.3 | — | — | — | — | — | |
| SAFormerNetwork=SAFormer-2-256, T=162025.12 | 98.3 | — | — | — | — | — | |
| SpikformerNetwork=Spikformer-2-256, Param(M)=2.57, T=162025.12 | 98.3 | — | — | — | — | — | |
| ESTSformerNetwork=EST-Sformer-2-256, T=162025.12 | 98.26 | — | — | — | — | — | |
| SpikformerType=Direct Training, T=162024.03 | 97.9 | — | — | — | — | — | |
| SEW-ResNetNetwork=SEW-ResNet, Param(M)=0.13, T=162025.12 | 97.9 | — | — | — | — | — | |
| S-TLLRModel=VGG-9, Time-steps=20, Batch size=162023.06 | 97.72 | — | — | 2.27 | 1.61 | — | |
| PLIFType=Direct Training, T=202024.03 | 97.57 | — | — | — | — | — | |
| CNN+EGRUunits=795, parameters=1.7M+3.1M, input size=128x128, data augmentation=true2022.06 | 97.3 | 77.6 | 72.7 | — | — | — | |
| CogniSNNNetwork=WS-RGA-7, Param(M)=0.13, T=162025.12 | 97.23 | — | — | — | — | — | |
| CNN+GRUunits=136, parameters=1.7M+0.4M, input size=128x128, data augmentation=true2022.06 | 97.15 | 0 | — | — | — | — | |
| teacher default-KDNetwork=ResNet18, Average timestep=162024.01 | 96.88 | — | — | — | — | — | |
| OTTModel=VGG-9, Time-steps=20, Batch size=162023.06 | 96.88 | — | — | — | — | — | |
| STBP-tdBNType=Direct Training, T=402024.03 | 96.87 | — | — | — | — | — | |
| CNN+EGRUunits=256, parameters=1.7M+1.0M, input size=128x128, data augmentation=true2022.06 | 96.8 | 73.4 | 55.3 | — | — | — | |
| CogniSNNNetwork=WS-RGA-7, Param(M)=0.13, T=82025.12 | 96.51 | — | — | — | — | — | |
| EGRUunits=1024, parameters=15.7M, input size=scaled, data augmentation=true2022.06 | 95.9 | 84.2 | 52.8 | — | — | — | |
| CogniSNNNetwork=ER-RGA-7, Param(M)=0.13, T=82025.12 | 95.81 | — | — | — | — | — | |
| BPTT (baseline)Model=VGG-9, Time-steps=20, Batch size=162023.06 | 95.58 | — | — | 6.06 | 16.13 | — | |
| DECOLLEModel=SNN (4 layers), Time-steps=1800, Batch size=722023.06 | 95.54 | — | — | — | — | — | |
| CogniSNNNetwork=WS-RGA-7, Param(M)=0.13, T=52025.12 | 95.41 | — | — | — | — | — | |
| SpikingGammaFrames=2000, Architecture=4-layer CNN2026.02 | 95.08 | — | — | — | — | 96.08 | |
| BPTT (baseline)Model=ResNet18, Time-steps=20, Batch size=162023.06 | 94.92 | — | — | 6.34 | 25.03 | — | |
| S-TLLRModel=ResNet18, Time-steps=20, Batch size=162023.06 | 94.92 | — | — | 2.27 | 2.5 | — | |
| SSNNNetwork=VGG-9, Average timestep=82024.01 | 94.91 | — | — | — | — | — | |
| SSNNNetwork=VGG-9, T=82025.12 | 94.91 | — | — | — | — | — | |
| GRUunits=1024, parameters=15.7M, input size=scaled, data augmentation=true2022.06 | 94.8 | 0 | — | — | — | — | |
| CogniSNNNetwork=ER-RGA-7, Param(M)=0.13, T=52025.12 | 94.78 | — | — | — | — | — | |
| SLAYERModel=SNN (8 layers), Time-steps=3002023.06 | 93.64 | — | — | — | — | — | |
| SpikingResformerType=Transfer Learning, T=102024.03 | 93.4 | — | — | — | — | — | |
| SpikingGammaFrames=2000, Architecture=2-layer FC2026.02 | 92.3 | — | — | — | — | 93.81 | |
| SSNNNetwork=VGG-9, Average timestep=52024.01 | 90.74 | — | — | — | — | — | |
| SSNNNetwork=VGG-9, T=52025.12 | 90.74 | — | — | — | — | — | |
| GRUunits=1024, parameters=15.7M, input size=scaled2022.06 | 88.1 | 0 | — | — | — | — | |
| LSTMunits=867, parameters=16.3M, input size=scaled2022.06 | 87.9 | 0 | — | — | — | — | |
| EGRUunits=1024, parameters=15.7M, input size=scaled2022.06 | 87.7 | 79.8 | 54.4 | — | — | — | |
| EGRUunits=512, parameters=5.5M, input size=scaled2022.06 | 86 | 76.1 | 45.7 | — | — | — | |
| MLFNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 85.77 | — | — | — | — | — | |
| EGRUunits=1024, parameters=110.1M, input size=128x1282022.06 | 85.7 | 77.3 | 64.6 | — | — | — | |
| PLIFNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 79.98 | — | — | — | — | — | |
| LIFTimesteps (T)=302026.02 | 79.92 | — | — | — | — | — | |
| UltraDPLIFTimesteps (T)=30, Type=Spatial (Diffusion PDE)2026.02 | 79.92 | — | — | — | — | — | |
| BackEISNNNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 79.86 | — | — | — | — | — | |
| SpikformerNetwork=Spikformer, Average timestep=5, Self-implementation=true2024.01 | 79.52 | — | — | — | — | — | |
| SpikformerNetwork=Spikformer-2-256, Param(M)=2.57, T=52025.12 | 79.52 | — | — | — | — | — | |
| STBP-tdBNNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 78.94 | — | — | — | — | — | |
| PLIFTimesteps (T)=302026.02 | 78.79 | — | — | — | — | — | |
| AdaLIFTimesteps (T)=302026.02 | 78.79 | — | — | — | — | — | |
| DSpikeTimesteps (T)=302026.02 | 78.79 | — | — | — | — | — | |
| TEBNNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 78.7 | — | — | — | — | — | |
| DSpike+Timesteps (T)=302026.02 | 78.41 | — | — | — | — | — | |
| UltraDLIFTimesteps (T)=30, Type=Spatial (Diffusion PDE)2026.02 | 78.41 | — | — | — | — | — | |
| FullPLIFTimesteps (T)=302026.02 | 77.27 | — | — | — | — | — | |
| UltraPLIFTimesteps (T)=30, Type=Temporal (LIF ODE)2026.02 | 75.76 | — | — | — | — | — | |
| UltraLIFTimesteps (T)=30, Type=Temporal (LIF ODE)2026.02 | 75 | — | — | — | — | — | |
| UltraLIFTimesteps (T)=10, Type=Temporal (LIF ODE)2026.02 | 69.32 | — | — | — | — | — | |
| UltraDLIFTimesteps (T)=10, Type=Spatial (Diffusion PDE)2026.02 | 69.32 | — | — | — | — | — | |
| PLIFTimesteps (T)=102026.02 | 68.94 | — | — | — | — | — | |
| AdaLIFTimesteps (T)=102026.02 | 68.94 | — | — | — | — | — | |
| UltraPLIFTimesteps (T)=10, Type=Temporal (LIF ODE)2026.02 | 68.94 | — | — | — | — | — | |
| FullPLIFTimesteps (T)=102026.02 | 68.56 | — | — | — | — | — | |
| UltraDPLIFTimesteps (T)=10, Type=Spatial (Diffusion PDE)2026.02 | 68.56 | — | — | — | — | — | |
| DSpikeTimesteps (T)=102026.02 | 67.42 | — | — | — | — | — | |
| LIFTimesteps (T)=102026.02 | 67.05 | — | — | — | — | — | |
| DSpike+Timesteps (T)=102026.02 | 66.67 | — | — | — | — | — | |
| UltraPLIFTimesteps (T)=1, Type=Temporal (LIF ODE)2026.02 | 60.23 | — | — | — | — | — | |
| UltraLIFTimesteps (T)=1, Type=Temporal (LIF ODE)2026.02 | 58.33 | — | — | — | — | — | |
| UltraDLIFTimesteps (T)=1, Type=Spatial (Diffusion PDE)2026.02 | 58.33 | — | — | — | — | — | |
| UltraDPLIFTimesteps (T)=1, Type=Spatial (Diffusion PDE)2026.02 | 58.33 | — | — | — | — | — | |
| LIFTimesteps (T)=12026.02 | 52.27 | — | — | — | — | — | |
| PLIFTimesteps (T)=12026.02 | 52.27 | — | — | — | — | — | |
| DSpikeTimesteps (T)=12026.02 | 51.14 | — | — | — | — | — | |
| DSpike+Timesteps (T)=12026.02 | 51.14 | — | — | — | — | — | |
| AdaLIFTimesteps (T)=12026.02 | 47.73 | — | — | — | — | — | |
| FullPLIFTimesteps (T)=12026.02 | 47.73 | — | — | — | — | — | |
| DECOLLEFrames=500, Architecture=4-layer CNN, CUBA LIF2026.02 | — | — | — | — | — | 95.54 | |
| FPTTFrames=1000, Architecture=2-layer FC, LTC2026.02 | — | — | — | — | — | 91.28 |