Partial Domain Adaptation on Office-31 (test)
98.98Accuracy (A -> W)BA³US
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| BA³USBackbone=ResNet-502020.03 | 98.98 | 100 | 98.73 | 99.36 | 94.82 | 94.99 | 97.81 | |
| MWPDABackbone=ResNet-502020.03 | 96.61 | 100 | 100 | 95.12 | 95.02 | 95.51 | 97.05 | |
| ETNBackbone=ResNet-502020.03 | 94.52 | 100 | 100 | 95.03 | 96.21 | 94.64 | 96.73 | |
| SANBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 93.9 | 99.3 | 99.4 | 94.3 | 94.2 | 88.7 | 95 | |
| SANBackbone=ResNet-502020.03 | 93.9 | 99.32 | 99.36 | 94.27 | 94.15 | 88.73 | 94.96 | |
| Ours (w/ BAA)Backbone=ResNet-502020.03 | 93.9 | 100 | 100 | 93.63 | 94.89 | 94.78 | 96.2 | |
| E-DANNBackbone=ResNet-502020.03 | 93.22 | 100 | 98.73 | 92.36 | 94.61 | 94.71 | 95.6 | |
| SSPDABackbone=ResNet-502020.03 | 91.52 | 92.88 | 98.94 | 90.87 | 90.61 | 94.36 | 93.2 | |
| DRCNLabel type=hard label, Backbone=ResNet-50, Pre-training=ImageNet2020.04 | 90.8 | 100 | 100 | 94.3 | 95.2 | 94.8 | 95.9 | |
| IWANBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 89.2 | 99.3 | 99.4 | 90.5 | 95.6 | 84.7 | 93.1 | |
| IWANBackbone=ResNet-502020.03 | 89.15 | 99.32 | 99.36 | 90.45 | 95.62 | 94.26 | 94.69 | |
| DRCNLabel type=soft label, Backbone=ResNet-50, Pre-training=ImageNet2020.04 | 88.5 | 100 | 100 | 86 | 95.6 | 95.8 | 94.3 | |
| DRCNBackbone=ResNet-502020.03 | 88.05 | 100 | 100 | 86 | 95.6 | 95.8 | 94.3 | |
| DEVBase Model=PADA, Validation Method=Deep Embedded Validation2026.06 | 87.8 | 100 | 100 | 82.94 | 92.84 | 95.23 | 93.15 | |
| Target Risk (Upper Bound)Base Model=PADA, Validation Method=Oracle (Target Risk)2026.06 | 87.8 | 100 | 100 | 83.59 | 93 | 95.66 | 93.34 | |
| PADABackbone=ResNet-502020.03 | 86.54 | 99.32 | 100 | 82.17 | 92.69 | 95.41 | 92.69 | |
| OriginalBase Model=PADA, Validation Method=Original report2026.06 | 86.54 | 99.32 | 100 | 82.17 | 92.69 | 95.41 | 92.69 | |
| PADABackbone=ResNet-50, Pre-training=ImageNet2020.04 | 86.5 | 99.3 | 100 | 82.2 | 92.7 | 95.4 | 92.7 | |
| IWCVBase Model=PADA, Validation Method=Importance Weighted Cross-Validation2026.06 | 82.38 | 97 | 96.42 | 78.96 | 89.16 | 92.23 | 89.36 | |
| CDAN+EBackbone=ResNet-502020.03 | 80.51 | 98.98 | 98.09 | 77.07 | 93.58 | 91.65 | 89.98 | |
| ADDABackbone=ResNet-502020.03 | 75.67 | 95.38 | 99.85 | 83.41 | 83.62 | 84.25 | 87.03 | |
| ResNet-50Backbone=ResNet-502020.03 | 75.59 | 96.27 | 98.09 | 83.44 | 83.92 | 84.97 | 87.05 | |
| RTNBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 75.3 | 97.1 | 98.3 | 66.9 | 85.6 | 85.7 | 84.8 | |
| LELBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 73.2 | 93.9 | 96.8 | 76.4 | 83.6 | 84.8 | 84.8 | |
| Source RiskBase Model=PADA, Validation Method=Source Risk2026.06 | 70.17 | 98.3 | 99.32 | 76.17 | 88.51 | 90.92 | 87.23 | |
| ResNetBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 54.5 | 94.6 | 94.3 | 65.6 | 73.2 | 71.7 | 75.6 | |
| RevGradBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 46.8 | 38.9 | 41.4 | 41.3 | 44.7 | 42.4 | 42.6 | |
| DANBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 46.4 | 53.8 | 58.6 | 42.7 | 65.7 | 65.3 | 55.4 | |
| ADDABackbone=ResNet-50, Pre-training=ImageNet2020.04 | 43.7 | 46.5 | 40.1 | 43.7 | 42.8 | 46 | 43.8 | |
| JANBackbone=ResNet-50, Pre-training=ImageNet2020.04 | 43.4 | 53.8 | 41.4 | 35.7 | 51 | 51.6 | 46.1 |