Feature Inversion on Dogs
51.7SSIMAdjoint Inversion
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Adjoint InversionBackbone=ResNet-502026.04 | 51.7 | 51.5 | |
| Adjoint InversionBackbone=ConvNeXt-Base2026.04 | 38.1 | 70 | |
| UpConvNetBackbone=ConvNeXt-Base2026.04 | 37.4 | 85.2 | |
| UpConvNetBackbone=ResNet-502026.04 | 36.4 | 84.4 | |
| Adjoint InversionBackbone=ResNet-182026.04 | 34.8 | 72.1 | |
| Adjoint InversionBackbone=DenseNet-1212026.04 | 34.8 | 63.5 | |
| UpConvNetBackbone=DenseNet-1212026.04 | 33.6 | 87 | |
| UpConvNetBackbone=ResNet-182026.04 | 32.3 | 91.4 | |
| MahendranBackbone=ConvNeXt-Base2026.04 | 14.1 | 81.3 | |
| MahendranBackbone=ResNet-502026.04 | 10.5 | 115.6 | |
| RobustBackbone=ConvNeXt-Base2026.04 | 10.1 | 99.6 | |
| MahendranBackbone=DenseNet-1212026.04 | 8.2 | 118.3 | |
| MahendranBackbone=ResNet-182026.04 | 6.5 | 131.9 | |
| RobustBackbone=ResNet-502026.04 | 6.2 | 120.8 | |
| Vanilla GradientBackbone=ConvNeXt-Base2026.04 | 5.2 | 125.7 | |
| RobustBackbone=DenseNet-1212026.04 | 5.2 | 123.9 | |
| Vanilla GradientBackbone=ResNet-502026.04 | 5.1 | 128 | |
| Vanilla GradientBackbone=DenseNet-1212026.04 | 4.7 | 123.6 | |
| RobustBackbone=ResNet-182026.04 | 4.4 | 131.4 | |
| Vanilla GradientBackbone=ResNet-182026.04 | 3.2 | 139.6 |