Feature Visualization Evaluation on ImageNet
70.64Zero-Shot Top-1 AccImageNet
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ImageNetArchitecture=DenseNet1212025.03 | 70.64 | — | — | — | 93.16 | |
| ImageNetArchitecture=ResNet502025.03 | 69.11 | — | — | — | 92.23 | |
| VITALArchitecture=ViT-L-162025.03 | 68.17 | 99.8 | 128.02 | 126.29 | 92.8 | |
| VITALArchitecture=ResNet50, Setup=train-set2025.03 | 66.62 | 99.9 | 58.79 | 58.79 | 92.56 | |
| ImageNetArchitecture=ViT-L-322025.03 | 65.83 | — | — | — | 90.03 | |
| ImageNetArchitecture=ConvNeXt base2025.03 | 65.66 | — | — | — | 89.8 | |
| ImageNetArchitecture=ViT-L-162025.03 | 64.78 | — | — | — | 89.31 | |
| VITALArchitecture=ConvNeXt base2025.03 | 63.53 | 99.97 | 88.63 | 3.92 | 90.3 | |
| VITALArchitecture=DenseNet1212025.03 | 58.7 | 99.93 | 79.4 | 0.27 | 86.93 | |
| VITALArchitecture=ViT-L-322025.03 | 55.97 | 89.6 | 174.31 | 147.33 | 85.47 | |
| PIIArchitecture=ViT-L-32, Batch size=52025.03 | 38.47 | 100 | 270.2 | 293.02 | 67.4 | |
| DeepInversionArchitecture=ResNet50, Batch size=642025.03 | 29.9 | 100 | 35.76 | 35.76 | 55.2 | |
| PIIArchitecture=ResNet50, Batch size=212025.03 | 17.53 | 100 | 241.54 | 241.54 | 38.93 | |
| MACOArchitecture=ViT-L-32, Resolution=2242025.03 | 17.53 | 24.87 | 280.77 | 2,318.9 | 37.23 | |
| DeepInversionArchitecture=ResNet50, Batch size=64, Multi-resolution optimization=true2025.03 | 17.2 | 50.47 | 176.35 | 176.35 | 40.43 | |
| PIIArchitecture=ViT-L-16, Batch size=22025.03 | 15.7 | 100 | 274.28 | 537.32 | 32.73 | |
| MACOArchitecture=ResNet50, Resolution=2242025.03 | 12.87 | 29.43 | 360.74 | 360.74 | 29.73 | |
| PIIArchitecture=DenseNet121, Batch size=242025.03 | 11 | 100 | 377.92 | 1.23 | 24 | |
| FourierArchitecture=ViT-L-32, Resolution=2242025.03 | 10.3 | 17.03 | 331.86 | 1,983.09 | 28.1 | |
| DeepInversionArchitecture=DenseNet121, Batch size=642025.03 | 10 | 100 | 93.26 | 0.2 | 25.47 | |
| MACOArchitecture=DenseNet121, Resolution=2242025.03 | 9.33 | 9.2 | 418.6 | 1.8 | 23.2 | |
| DeepInversionArchitecture=DenseNet121, Batch size=64, Multi-resolution optimization=true2025.03 | 7.23 | 31.3 | 186.16 | 0.83 | 20.03 | |
| MACOArchitecture=ConvNeXt base, Resolution=2242025.03 | 7.2 | 66.07 | 369.64 | 62.55 | 19.77 | |
| FourierArchitecture=ResNet50, Resolution=2242025.03 | 6.73 | 21.3 | 422.44 | 422.44 | 18.27 | |
| DeepInversionArchitecture=ResNet50, Batch size=1, Multi-resolution optimization=true2025.03 | 6.3 | 100 | 121.94 | 121.94 | 16.43 | |
| FourierArchitecture=DenseNet121, Resolution=2242025.03 | 4.87 | 15.53 | 409.89 | 1.63 | 12.17 | |
| DeepInversionArchitecture=ResNet50, Batch size=12025.03 | 4.73 | 100 | 123.77 | 123.77 | 12.63 | |
| MACOArchitecture=ViT-L-16, Resolution=2242025.03 | 3.93 | 44.33 | 371.54 | 946.96 | 10.57 | |
| FourierArchitecture=ConvNeXt base, Resolution=2242025.03 | 2.77 | 60.07 | 453.91 | 59.6 | 8.3 | |
| PIIArchitecture=ConvNeXt base, Batch size=162025.03 | 1.97 | 100 | 405.5 | 92.37 | 6.47 | |
| FourierArchitecture=ViT-L-16, Resolution=2242025.03 | 1.67 | 25.3 | 447.56 | 990.51 | 5.13 | |
| MACOArchitecture=ResNet50, Resolution=10242025.03 | 1.6 | 2.1 | 494.57 | 494.57 | 5.67 | |
| MACOArchitecture=DenseNet121, Resolution=10242025.03 | 1.43 | 1.6 | 475.39 | 1.98 | 5.03 | |
| MACOArchitecture=ConvNeXt base, Resolution=10242025.03 | 1.07 | 21.07 | 495.69 | 97.73 | 4.73 | |
| FourierArchitecture=ResNet50, Resolution=10242025.03 | 0.97 | 3.43 | 430.58 | 430.58 | 3.57 | |
| FourierArchitecture=DenseNet121, Resolution=10242025.03 | 0.9 | 1.8 | 437.18 | 1.88 | 3.33 | |
| FourierArchitecture=ConvNeXt base, Resolution=10242025.03 | 0.6 | 14.27 | 529.33 | 77.08 | 2.37 |