Image Classification on CIFAR10-DVS (test)
88.4AccuracyKvLIF
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| KvLIFNetwork Architecture=VGGSNN, Timestep=10, implementation_note=neuron model modification solely2026.03 | 88.4 | — | |
| AR-LIFNetwork Architecture=VGGSNN, Timestep=162026.03 | 87.9 | — | |
| VGG (ours)Parameter (M)=9.30, Time Steps=102026.05 | 87 | — | |
| CLIFNetwork Architecture=VGGSNN, Timestep=102026.03 | 86.1 | — | |
| PSNNetwork Architecture=VGGSNN, Timestep=102026.03 | 85.9 | — | |
| Spikformer (ours)Parameter (M)=2.57, Time Steps=102026.05 | 85.07 | — | |
| GETInput Type=Token, Params (M)=4.52023.10 | 84.8 | — | |
| QKFormerParameter (M)=1.50, Time Steps=42026.05 | 84 | — | |
| SWformerParam. (M)=2.05, T=162026.05 | 83.9 | — | |
| STAttenT=162026.05 | 83.9 | — | |
| TETModel=VGGSNN, Training Protocol=Direct Training of SNNs2023.04 | 83.1 | 10 | |
| SEENN-IModel=VGG-SNN, Training Protocol=Direct Training of SNNs, Timesteps=5.172023.04 | 82.7 | 5.17 | |
| SEENN-IIModel=VGG-SNN, Training Protocol=Direct Training of SNNs2023.04 | 82.6 | 4.49 | |
| UncertParam. (M)=1.50, T=102026.05 | 82.5 | — | |
| QKFormerParam. (M)=1.50, T=102026.05 | 82.3 | — | |
| SpikFormer + SEMMTime Steps=162026.05 | 82.1 | — | |
| SpikformerParam. (M)=2.57, T=162026.05 | 80.9 | — | |
| SpikformerParameter (M)=2.57, Time Steps=162026.05 | 80.9 | — | |
| ANN-Guided DistillationModel=ResNet-19, Timesteps=42025.03 | 80.54 | — | |
| CogniSNNNetwork=WS-RGA-7, Param(M)=1.51, T=82025.12 | 80.52 | — | |
| CMLNetwork=Spikformer-4-384, T=102025.12 | 80.5 | — | |
| CogniSNNNetwork=ER-RGA-7, Param(M)=1.51, T=82025.12 | 80.5 | — | |
| SAKDModel=ResNet-19, Timesteps=42025.03 | 80.3 | — | |
| ENOFModel=ResNet19, Timesteps=102025.03 | 80.1 | — | |
| S-TransformerParam. (M)=2.57, T=162026.05 | 80 | — | |
| CogniSNNNetwork=ER-RGA-7, Param(M)=1.51, T=52025.12 | 79.8 | — | |
| STSANetwork=STS-Transformer-1-256, T=162025.12 | 79.6 | — | |
| KvLIFNetwork Architecture=VGG-11, Timestep=102026.03 | 79.5 | — | |
| CogniSNNNetwork=WS-RGA-7, Param(M)=1.51, T=52025.12 | 79 | — | |
| CLIFNetwork Architecture=VGG-11, Timestep=162026.03 | 79 | — | |
| SpikformerNetwork=Spikformer-2-256, Param(M)=2.57, T=102025.12 | 78.9 | — | |
| ILIFNetwork Architecture=VGG-11, Timestep=102026.03 | 78.6 | — | |
| SSNNNetwork=VGG-9, Average timestep=82024.01 | 78.57 | — | |
| SSNNNetwork=VGG-9, T=82025.12 | 78.57 | — | |
| STSANetwork=STS-Transformer-1-256, T=102025.12 | 78.28 | — | |
| Nested-TInput Type=Token, Params (M)=4.22023.10 | 78.1 | — | |
| GLIFModel=7B-wideNet, Timesteps=162025.03 | 78.1 | — | |
| GLIFNetwork Architecture=7B-wideNet, Timestep=162026.03 | 78.1 | — | |
| SEENN-IModel=VGG-SNN, Training Protocol=Direct Training of SNNs, Timesteps=2.532023.04 | 77.6 | 2.53 | |
| SLSSNNNetwork=VGG-16, Average timestep=82024.01 | 77.5 | — | |
| TETNetwork=VGGSNN, Average timestep=102024.01 | 77.33 | — | |
| DSRNetwork=VGG-11, Average timestep=20, Data augmentation=true2024.01 | 77.27 | — | |
| Swin-T v2Input Type=Token, Params (M)=6.92023.10 | 77.2 | — | |
| SLTTNetwork=VGG-11, Average timestep=102024.01 | 77.17 | — | |
| RateBPModel=VGG-11, Timesteps=102025.03 | 76.96 | — | |
| GLIFNetwork=7B-wideNet, Average timestep=162024.01 | 76.8 | — | |
| MVF-NetPre-training=Pretrained on ImageNet [9], Backbone=ResNet-34 [21]2021.06 | 76.2 | — | |
| DspikeModel=ResNet-18, Training Protocol=Direct Training of SNNs2023.04 | 75.4 | 10 | |
| DspikeModel=ResNet-18, Timesteps=102025.03 | 75.4 | — | |
| DspikeNetwork Architecture=ResNet-18, Timestep=102026.03 | 75.4 | — | |
| ESTPre-training=Pretrained on ImageNet [9], Backbone=ResNet-34 [21]2021.06 | 74.9 | — | |
| FC100-PLIF-APk10s10T=20, Parameters=17.4M2021.02 | 74.8 | — | |
| PLIFModel=VGGSNN, Training Protocol=Direct Training of SNNs2023.04 | 74.8 | 20 | |
| PLIFInput Type=Spike, Params (M)=4.7 - 17.12023.10 | 74.8 | — | |
| PLIFNetwork Architecture=PLIF-Net, Timestep=202026.03 | 74.8 | — | |
| Wide-7B-NetT=16, Parameters=1.19M, SEW Block=true2021.02 | 74.4 | — | |
| SSNNNetwork=VGG-9, Average timestep=52024.01 | 73.63 | — | |
| SSNNNetwork=VGG-9, T=52025.12 | 73.63 | — | |
| M-LSTMPre-training=Pretrained on ImageNet [9], Backbone=ResNet-34 [21]2021.06 | 73 | — | |
| Real SpikeModel=ResNet-19, Timesteps=102025.03 | 72.85 | — | |
| AutoSNNNetwork=AutoSNN, Average timestep=8, Data augmentation=true2024.01 | 72.5 | — | |
| RecDis-SNNModel=ResNet19, Timesteps=102025.03 | 72.42 | — | |
| RecDis-SNNModel=ResNet-19, Training Protocol=Direct Training of SNNs2023.04 | 72.4 | 10 | |
| LIAF-NetModel=LIAF-Net, Timesteps=102025.03 | 71.7 | — | |
| Wide-7B-NetT=8, Parameters=1.19M, SEW Block=true2021.02 | 70.2 | — | |
| SEW-ResNetNetwork=SEW-ResNet, Param(M)=1.19, T=82025.12 | 70.2 | — | |
| SpikformerNetwork=Spikformer, Average timestep=5, Self-implementation=true, Data augmentation=true2024.01 | 68.55 | — | |
| SpikformerNetwork=Spikformer-2-256, Param(M)=2.57, T=52025.12 | 68.55 | — | |
| EvSInput Type=Graph2023.10 | 68 | — | |
| Spiking ResNet-19T=10, Parameters=11.18M, td-BN=true2021.02 | 67.8 | — | |
| tdBNModel=ResNet-19, Training Protocol=Direct Training of SNNs2023.04 | 67.8 | 10 | |
| STBP-tdBNModel=ResNet-19, Timesteps=102025.03 | 67.8 | — | |
| TEBNNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 67.7 | — | |
| MLFNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 67.07 | — | |
| MLFNetwork=VGG-9, T=52025.12 | 67.07 | — | |
| EV-VGCNNPre-training=Without pretraining2021.06 | 67 | — | |
| EV-VGCNN (Ours)2021.06 | 67 | — | |
| STBP-tdBNNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 66.93 | — | |
| RolloutModel=DenseNet, Timesteps=102025.03 | 66.8 | — | |
| PLIFNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 66.77 | — | |
| AsyNetPre-training=Without pretraining2021.06 | 66.3 | — | |
| VMV-GCNInput Type=Voxel, Params (M)=0.92023.10 | 66.3 | — | |
| EV-VGCNN (Ours)Variation=w/ SFRL2021.06 | 65.2 | — | |
| EV-VGCNNInput Type=Voxel, Params (M)=0.82023.10 | 65.1 | — | |
| Wide-7B-NetT=4, Parameters=1.19M, SEW Block=true2021.02 | 64.8 | — | |
| ESTPre-training=Without pretraining, Backbone=ResNet-34 [21]2021.06 | 63.4 | — | |
| M-LSTMPre-training=Without pretraining, Backbone=ResNet-34 [21]2021.06 | 63.1 | — | |
| BackEISNNNetwork=VGG-9, Average timestep=5, Self-implementation=true2024.01 | 62.9 | — | |
| AMAEInput Type=Frame, Params (M)=21.82023.10 | 62 | — | |
| DT-SNNInput Type=Spike2023.10 | 60.5 | — | |
| MVF-NetPre-training=Without pretraining, Backbone=ResNet-34 [21]2021.06 | 59.9 | — | |
| MVF-NetInput Type=Frame, Params (M)=33.52023.10 | 55.8 | — | |
| RG-CNNs2021.06 | 54 | — | |
| RG-CNNsInput Type=Voxel, Params (M)=19.52023.10 | 54 | — | |
| PointNet++Input representation=Proposed representation2021.06 | 53.3 | — | |
| HATS2021.06 | 52.4 | — | |
| HATSInput Type=Frame2023.10 | 52.4 | — | |
| PointNet++Input representation=Point-wise2021.06 | 46.5 | — | |
| HOTS2021.06 | 27.1 | — | |
| EventNet2021.06 | 17.1 | — |