Multi-label Classification on reBEN Random 1.0 (test)
71.75mAPGT w/ BCE
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| GT w/ BCEMethod category=Ground-truth Training, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 71.75 | 4.264 | 3.428 | 93.09 | 66.63 | 72.07 | 63.24 | |
| IUNMethod category=Ground-truth Training, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 70.46 | 4.331 | 3.608 | 91.98 | 56 | 84.75 | 44.66 | |
| AdaGCMethod category=Proposed, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 67.85 | 4.553 | 4.165 | 92.49 | 64.46 | 70.43 | 63.22 | |
| RCMLMethod category=RS-specific Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 64.91 | 4.585 | 4.467 | 91.24 | 52.9 | 78.57 | 42.42 | |
| ELRMethod category=General Label Noise Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 62.4 | 4.86 | 4.89 | 91.45 | 63.97 | 62.63 | 68.26 | |
| LL-CtMethod category=General Label Noise Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 62.35 | 4.722 | 4.856 | 91.76 | 54.87 | 76.27 | 46.41 | |
| Dist-PUMethod category=PUL Methods, Backbone=ResNet-34, Batch size=512, Epochs=70, Additional a-priori information=true2025.10 | 61.98 | 4.86 | 5.111 | 90.67 | 61.49 | 59.11 | 68.83 | |
| BoostLUMethod category=General Label Noise Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 61.77 | 4.662 | 4.78 | 91.59 | 55.41 | 74.8 | 46.32 | |
| LL-RMethod category=General Label Noise Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 59.7 | 4.688 | 5.054 | 90.85 | 53.91 | 70.44 | 45.45 | |
| WANMethod category=Baseline Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 59.43 | 4.66 | 5.089 | 90.16 | 60.15 | 55.96 | 66.14 | |
| MIMEMethod category=SPML Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 58.64 | 4.894 | 5.632 | 90.66 | 55.13 | 62.4 | 50.39 | |
| LACMethod category=SPML Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 57.88 | 4.801 | 5.384 | 87.33 | 28.97 | 72.29 | 19.85 | |
| GRMethod category=SPML Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 57.38 | 4.914 | 5.855 | 89.77 | 56.79 | 57.91 | 56.06 | |
| LL-CpMethod category=General Label Noise Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 57.27 | 4.983 | 6.068 | 89.6 | 49.35 | 65.47 | 43.67 | |
| ROLEMethod category=SPML Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 55.73 | 6.479 | 6.424 | 91.18 | 59.65 | 61.9 | 58.36 | |
| ANMethod category=Baseline Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 54.97 | 4.849 | 5.847 | 87.13 | 28.15 | 72.93 | 19 | |
| LCRMethod category=RS-specific Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 54.56 | 4.864 | 5.912 | 87.16 | 28.52 | 73.82 | 19.31 | |
| AN-LSMethod category=Baseline Methods, Backbone=ResNet-34, Batch size=512, Epochs=702025.10 | 54.4 | 5.375 | 7.051 | 86.9 | 25.47 | 75.9 | 17.07 | |
| EPRMethod category=Baseline Methods, Backbone=ResNet-34, Batch size=512, Epochs=70, Additional a-priori information=true2025.10 | 50.74 | 5.206 | 6.825 | 80.13 | 51.22 | 38.25 | 85.59 |