Image Classification on ImageNet 1K (Top-1 Accuracy and GFLOPs)
89.2Top-1 AccuracyDreamNet-4
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DreamNet-4Type=DreamNet, Input Size=224^2, #Params=212.03, FLOPs=118.42024.09 | 89.2 | — | |
| DreamNet-3MAE-lType=DreamNet, Input Size=224^2, #Params=209.90, FLOPs=149.42024.09 | 88.9 | — | |
| DreamNet-3Type=DreamNet, Input Size=224^2, #Params=209.12, FLOPs=83.52024.09 | 87.8 | — | |
| DreamNet-3MAE-bType=DreamNet, Input Size=224^2, #Params=209.12, FLOPs=83.52024.09 | 87.8 | — | |
| SleepNet-3ViT-lType=SleepNet, Input Size=224^2, #Params=13.40, FLOPs=371.52024.09 | 86.4 | — | |
| SleepNet-3Type=SleepNet, Input Size=224^2, #Params=10.07, FLOPs=60.62024.09 | 85.9 | — | |
| SleepNet-3ViT-bType=SleepNet, Input Size=224^2, #Params=10.07, FLOPs=107.32024.09 | 85.9 | — | |
| EfficientNetV2-LType=Conv only, Input Size=480^2, #Params=121.32, FLOPs=53.02024.09 | 85.7 | — | |
| ViTl-ACNType=Augmentation +Trans, Input Size=384^2, #Params=490.11, FLOPs=41.92024.09 | 85.7 | — | |
| EfficientNet-B7Type=Conv only, Input Size=600^2, #Params=66.44, FLOPs=37.02024.09 | 84.7 | — | |
| CoAtNet-3Type=Conv + Trans, Input Size=224^2, #Params=168.12, FLOPs=34.72024.09 | 84.5 | — | |
| CoAtNet-2Type=Conv + Trans, Input Size=224^2, #Params=75.14, FLOPs=15.72024.09 | 84.1 | — | |
| SleepNet-3MAE-lType=SleepNet, Input Size=224^2, #Params=10.09, FLOPs=126.52024.09 | 84.1 | — | |
| SleepNet-4Type=SleepNet, Input Size=224^2, #Params=10.14, FLOPs=80.72024.09 | 83.9 | — | |
| MAEl-MASType=Augmentation +Trans, Input Size=224^2, #Params=551.42, FLOPs=29.92024.09 | 83.6 | — | |
| CvT-21Type=Conv + Trans, Input Size=384^2, #Params=32.41, FLOPs=24.92024.09 | 83.3 | — | |
| MAElargeType=Transformer, Input Size=224^2, #Params=304.18, FLOPs=61.92024.09 | 83.1 | — | |
| ViTlargeType=Transformer, Input Size=384^2, #Params=307.12, FLOPs=190.72024.09 | 83 | — | |
| MAEbaseType=Transformer, Input Size=224^2, #Params=86.43, FLOPs=17.62024.09 | 82.6 | — | |
| SleepNet-3MAE-bType=SleepNet, Input Size=224^2, #Params=10.07, FLOPs=60.62024.09 | 81.9 | — | |
| BaselineModel=DeiT-B, Params (M)=86, Training-free=True2025.05 | 81.8 | 17.6 | |
| EViTModel=DeiT-B, Params (M)=86, Training-free=True2025.05 | 81.8 | 15.9 | |
| ToMeModel=DeiT-B, Params (M)=86, Training-free=True2025.05 | 81.8 | 17 | |
| MCTFModel=DeiT-B, Params (M)=86, Training-free=True2025.05 | 81.8 | 16.2 | |
| ATMModel=DeiT-B, Params (M)=86, Training-free=True2025.05 | 81.8 | 13.9 | |
| ResNet101Type=Conv only, Input Size=224^2, #Params=45.01, FLOPs=7.62024.09 | 80.8 | — | |
| BaselineModel=DeiT-S, Params (M)=22, Training-free=True2025.05 | 79.8 | 4.6 | |
| EViTModel=DeiT-S, Params (M)=22, Training-free=True2025.05 | 79.8 | 4.2 | |
| ToMeModel=DeiT-S, Params (M)=22, Training-free=True2025.05 | 79.8 | 4.2 | |
| MCTFModel=DeiT-S, Params (M)=22, Training-free=True2025.05 | 79.8 | 3.6 | |
| Zero-TPModel=DeiT-S, Params (M)=22, Training-free=True2025.05 | 79.8 | 4 | |
| ATMModel=DeiT-S, Params (M)=22, Training-free=True2025.05 | 79.8 | 3.2 | |
| ViTbaseType=Transformer, Input Size=224^2, #Params=86.45, FLOPs=55.42024.09 | 79.4 | — | |
| ResNet-50# Parameters=25.56M2026.05 | 76.89 | — | |
| WONN (Ch = 256 → 256)S(θi)&I(θi) configuration type=as MLPs, Channel configuration (Ch)=256 → 256, # Parameters=12.28M2026.05 | 76.78 | — | |
| ResNet50Type=Conv only, Input Size=224^2, #Params=25.12, FLOPs=3.82024.09 | 76 | — | |
| ViT-B-16†# Parameters=86.57M2026.05 | 75.85 | — | |
| ViT-S-16†# Parameters=22.05M2026.05 | 75.54 | — | |
| WONN (Ch = 64 → 256)S(θi)&I(θi) configuration type=as MLPs, Channel configuration (Ch)=64 → 256, # Parameters=7.79M2026.05 | 74.84 | — | |
| BaselineModel=DeiT-T, Params (M)=5, Training-free=True2025.05 | 72.1 | 1.3 | |
| EViTModel=DeiT-T, Params (M)=5, Training-free=True2025.05 | 72.1 | 1.2 | |
| ToMeModel=DeiT-T, Params (M)=5, Training-free=True2025.05 | 72.1 | 1.2 | |
| MCTFModel=DeiT-T, Params (M)=5, Training-free=True2025.05 | 72.1 | 1.2 | |
| ATMModel=DeiT-T, Params (M)=5, Training-free=True2025.05 | 72.1 | 0.9 | |
| DreamNet-2Type=DreamNet, Input Size=224^2, #Params=208.31, FLOPs=52.32024.09 | 71.6 | — | |
| ResNet-18# Parameters=11.69M2026.05 | 69.73 | — | |
| ResNet18Type=Conv only, Input Size=224^2, #Params=11.23, FLOPs=1.82024.09 | 69.7 | — | |
| AKOrNattn# Parameters=4.85M2026.05 | 67.45 | — | |
| SleepNet-2Type=SleepNet, Input Size=224^2, #Params=10.01, FLOPs=40.52024.09 | 63 | — |