Image Classification on Rotated-MNIST (Rotation Robustness)
98.2Mean AccuracyARM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| ARMBackbone=MNIST-ConvNet2021.10 | 98.2 | — | — | — | — | — | — | |
| ARMArchitecture=MNIST ConvNet2023.02 | 98.1 | 95.9 | 99 | 98.8 | 98.9 | 99.1 | 96.7 | |
| DANN+ELSArchitecture=MNIST ConvNet2023.02 | 98.1 | 96.3 | 98.7 | 98.9 | 99.1 | 98.7 | 96.9 | |
| ERMBackbone=MNIST-ConvNet2021.10 | 98 | — | — | — | — | — | — | |
| GroupDROBackbone=MNIST-ConvNet2021.10 | 98 | — | — | — | — | — | — | |
| MixupBackbone=MNIST-ConvNet2021.10 | 98 | — | — | — | — | — | — | |
| CORALBackbone=MNIST-ConvNet2021.10 | 98 | — | — | — | — | — | — | |
| SagNetsBackbone=MNIST-ConvNet2021.10 | 98 | — | — | — | — | — | — | |
| mDSDIBackbone=MNIST-ConvNet2021.10 | 98 | — | — | — | — | — | — | |
| MLDGBackbone=MNIST-ConvNet2021.10 | 97.9 | — | — | — | — | — | — | |
| MMDBackbone=MNIST-ConvNet2021.10 | 97.9 | — | — | — | — | — | — | |
| CDANNBackbone=MNIST-ConvNet2021.10 | 97.9 | — | — | — | — | — | — | |
| MTLBackbone=MNIST-ConvNet2021.10 | 97.9 | — | — | — | — | — | — | |
| VREXBackbone=MNIST-ConvNet2021.10 | 97.9 | — | — | — | — | — | — | |
| DANNArchitecture=MNIST ConvNet2023.02 | 97.9 | 95.9 | 98.6 | 98.7 | 99 | 98.7 | 96.5 | |
| PerfMatchArchitecture=LeNet2020.06 | 97.8 | 96.5 | 99.1 | 99.2 | 98.6 | 98.6 | 94.9 | |
| DANNBackbone=MNIST-ConvNet2021.10 | 97.8 | — | — | — | — | — | — | |
| ERMArchitecture=MNIST ConvNet2023.02 | 97.8 | 95.3 | 98.7 | 98.9 | 98.7 | 98.9 | 96.2 | |
| IRMBackbone=MNIST-ConvNet2021.10 | 97.7 | — | — | — | — | — | — | |
| RSCBackbone=MNIST-ConvNet2021.10 | 97.6 | — | — | — | — | — | — | |
| IRMArchitecture=MNIST ConvNet2023.02 | 97.5 | 94.9 | 98.7 | 98.6 | 98.6 | 98.7 | 95.2 | |
| MatchDGArchitecture=LeNet2020.06 | 97.4 | 93 | 99.5 | 99.9 | 99.4 | 99.7 | 93.3 | |
| DIVAArchitecture=LeNet2020.06 | 97.2 | 93.5 | 99.3 | 99.1 | 99.2 | 99.3 | 93 | |
| RandMatchArchitecture=LeNet2020.06 | 97.1 | 91 | 99.7 | 99.6 | 99.4 | 99.7 | 93.1 | |
| Feature-Critic-MLP2019.01 | 96.39 | 89.23 | 99.68 | 99.2 | 99.24 | 99.53 | 91.44 | |
| Feature-Critic-Flatten2019.01 | 96.04 | 87.04 | 99.53 | 99.41 | 99.52 | 99.23 | 91.52 | |
| CrossGradArchitecture=LeNet2020.06 | 95.3 | 88.3 | 98.6 | 98 | 97.7 | 97.7 | 91.4 | |
| LabelGradArchitecture=LeNet2020.06 | 95.2 | 89.7 | 97.8 | 98 | 97.1 | 96.6 | 92.1 | |
| Reptile2019.01 | 95.15 | 87.78 | 99.44 | 98.42 | 98.8 | 99.03 | 87.42 | |
| AGG2019.01 | 95.14 | 86.42 | 98.61 | 99.19 | 98.22 | 99.48 | 88.92 | |
| MetaReg2019.01 | 94.97 | 85.7 | 98.87 | 98.32 | 98.58 | 98.93 | 89.44 | |
| CrossGrad2019.01 | 94.93 | 86.03 | 98.92 | 98.6 | 98.39 | 98.68 | 88.94 | |
| DANArchitecture=LeNet2020.06 | 94.3 | 86.7 | 98 | 97.8 | 97.4 | 96.9 | 89.1 | |
| ERMArchitecture=LeNet2020.06 | 94.1 | 88.2 | 98.6 | 97.7 | 97.5 | 97 | 85.6 | |
| Source-Specific Nets2018.06 | 92 | 85.6 | 95 | 95.6 | 95.5 | 95.9 | 84.3 | |
| [15]2018.06 | 89.1 | 84.6 | 95.6 | 94.6 | 82.9 | 94.8 | 82.1 | |
| CCSAArchitecture=LeNet2020.06 | 89.1 | 84.6 | 95.6 | 94.6 | 82.9 | 94.8 | 82.1 | |
| D-MTAEArchitecture=LeNet2020.06 | 87.6 | 82.5 | 96.3 | 93.4 | 78.6 | 94.2 | 80.5 | |
| [5]2018.06 | 87.5 | 82.5 | 96.3 | 93.4 | 78.6 | 94.2 | 80.5 | |
| [17]2018.06 | 85.5 | 72.1 | 95.3 | 92.6 | 81.5 | 92.7 | 79.3 |