Domain Adaptation on Office31 standard (test)
93.37Standard Accuracy (A->D)PL
Evaluation results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PLepsilon=8/255, AutoAttack (AA) evaluation protocol=true2025.05 | 93.37 | 93.37 | 94.72 | 94.34 | 73.59 | 71.81 | 98.49 | 98.37 | 74.26 | 72.63 | 99.8 | 99.6 | 89.04 | 88.35 | |
| TAROTepsilon=8/255, AutoAttack (AA) evaluation protocol=true2025.05 | 93.37 | 92.97 | 94.47 | 94.47 | 76.32 | 75.19 | 98.62 | 98.49 | 72.74 | 71.64 | 100 | 100 | 90.45 | 90.04 | |
| TAROTepsilon=4/255, Robust accuracy evaluation=AutoAttack2025.05 | 93.37 | 93.17 | 93.84 | 93.59 | 75.22 | 74.55 | 98.49 | 98.49 | 74.51 | 73.55 | 100 | 100 | 91 | 90.72 | |
| SROUDAepsilon=4/255, Robust accuracy evaluation=AutoAttack2025.05 | 92.97 | 92.77 | 95.22 | 94.21 | 74.62 | 65.74 | 98.74 | 98.74 | 66.45 | 64.57 | 100 | 100 | 88 | 86.01 | |
| rRevGrad+CATBackbone=ResNet-502019.03 | 90.8 | — | 94.4 | — | 72.2 | — | 98 | — | 70.2 | — | 100 | — | 87.6 | — | |
| CATBackbone=ResNet-502019.03 | 90.6 | — | 91.1 | — | 70.4 | — | 98.6 | — | 66.5 | — | 99.6 | — | 86.1 | — | |
| SROUDAepsilon=8/255, AutoAttack (AA) evaluation protocol=true2025.05 | 89.96 | 85.54 | 91.57 | 90.57 | 49.38 | 22.36 | 97.99 | 90.31 | 71.92 | 65.71 | 98.59 | 97.99 | 83.24 | 75.41 | |
| PLepsilon=4/255, Robust accuracy evaluation=AutoAttack2025.05 | 89.56 | 89.56 | 93.46 | 93.33 | 75.04 | 74.55 | 98.49 | 98.49 | 72.7 | 72.7 | 100 | 100 | 88.21 | 88.11 | |
| JAN+CATBackbone=ResNet-502019.03 | 88.1 | — | 94 | — | 68.9 | — | 96.6 | — | 69.4 | — | 100 | — | 86.2 | — | |
| GenToAdaptBackbone=ResNet-502019.03 | 87.7 | — | 89.5 | — | 72.8 | — | 97.9 | — | 71.4 | — | 99.8 | — | 86.5 | — | |
| SimNetBackbone=ResNet-502019.03 | 85.3 | — | 88.6 | — | 73.4 | — | 98.2 | — | 71.8 | — | 99.7 | — | 86.2 | — | |
| JANBackbone=ResNet-502019.03 | 84.7 | — | 85.4 | — | 68.6 | — | 97.4 | — | 70 | — | 99.8 | — | 84.3 | — | |
| RFAepsilon=4/255, Robust accuracy evaluation=AutoAttack2025.05 | 83.53 | 78.11 | 81.89 | 72.58 | 61.38 | 54.03 | 97.48 | 96.73 | 63.44 | 56.12 | 100 | 99.2 | 81.29 | 76.13 | |
| RevGradBackbone=ResNet-502019.03 | 79.4 | — | 82 | — | 68.2 | — | 96.9 | — | 67.4 | — | 99.1 | — | 82.2 | — | |
| DANBackbone=ResNet-502019.03 | 78.6 | — | 80.5 | — | 63.6 | — | 97.1 | — | 62.8 | — | 99.6 | — | 80.4 | — | |
| RFAepsilon=8/255, AutoAttack (AA) evaluation protocol=true2025.05 | 78.51 | 45.18 | 73.84 | 33.08 | 62.3 | 46.57 | 98.24 | 79.87 | 61.02 | 43.95 | 99.2 | 81.53 | 78.85 | 55.03 | |
| rRevGrad+CATBackbone=AlexNet2019.03 | 76.4 | — | 80.7 | — | 63.7 | — | 97.6 | — | 62.2 | — | 100 | — | 80.1 | — | |
| CATBackbone=AlexNet2019.03 | 74.7 | — | 77.4 | — | 63.4 | — | 97.4 | — | 60.8 | — | 99.9 | — | 78.9 | — | |
| MSTNBackbone=AlexNet2019.03 | 74.5 | — | 80.5 | — | 62.5 | — | 96.9 | — | 60 | — | 99.9 | — | 79.1 | — | |
| JAN+CATBackbone=AlexNet2019.03 | 74.5 | — | 78.4 | — | 63.6 | — | 97.2 | — | 61.2 | — | 100 | — | 79.2 | — | |
| RevGradBackbone=AlexNet2019.03 | 72.3 | — | 73 | — | 53.4 | — | 96.4 | — | 51.2 | — | 99.2 | — | 74.3 | — | |
| ARTUDAepsilon=4/255, Robust accuracy evaluation=AutoAttack2025.05 | 71.89 | 71.69 | 73.71 | 73.33 | 57.93 | 52.25 | 93.21 | 93.08 | 58.93 | 52.68 | 98.39 | 97.99 | 75.68 | 73.5 | |
| JANBackbone=AlexNet2019.03 | 71.8 | — | 74.9 | — | 58.3 | — | 96.6 | — | 55 | — | 99.5 | — | 76 | — | |
| ResNet-50Backbone=ResNet-50, Note=Source Only2019.03 | 68.9 | — | 68.4 | — | 62.5 | — | 96.7 | — | 60.7 | — | 99.3 | — | 76.1 | — | |
| DRCNBackbone=AlexNet2019.03 | 66.8 | — | 68.7 | — | 56 | — | 96.4 | — | 54.9 | — | 99 | — | 73.6 | — | |
| DDCBackbone=AlexNet2019.03 | 64.4 | — | 61.8 | — | 52.1 | — | 95 | — | 52.2 | — | 98.5 | — | 70.6 | — | |
| AlexNetBackbone=AlexNet, Note=Source Only2019.03 | 63.8 | — | 61.6 | — | 51.1 | — | 95.4 | — | 49.8 | — | 99 | — | 70.1 | — | |
| ARTUDAepsilon=8/255, AutoAttack (AA) evaluation protocol=true2025.05 | 47.79 | 45.58 | 47.67 | 45.16 | 42.88 | 33.12 | 88.81 | 86.54 | 59.99 | 36.74 | 94.18 | 91.57 | 63.55 | 56.45 |