Open-Set Recognition on Caltech-UCSD-Birds (CUB) Easy split 42 (test)
90.8Closed-set AccuracySpHOR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SpHORBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=MaxLogit2025.03 | 90.8 | 91.72 | 85.72 | |
| SpHORBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=KNN2025.03 | 90.8 | 90.01 | 84.28 | |
| SpHORBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=NNGuide2025.03 | 90.8 | 89.02 | 83.14 | |
| SpHORBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=PostMax2025.03 | 90.8 | 91.35 | 85.35 | |
| SpHORBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=MaxLogit, Training Strategy=w/o ROrtho2025.03 | 90.3 | 91.1 | 85.09 | |
| MLSBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=MaxLogit, Training Strategy=Mixup2025.03 | 88.3 | 86.19 | 78.64 | |
| MLSBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=PostMax, Training Strategy=Mixup2025.03 | 88.3 | 84.34 | 77.03 | |
| MLSBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=NNGuide, Training Strategy=Mixup2025.03 | 88.3 | 86.56 | 77.88 | |
| MLSBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=KNN, Training Strategy=Mixup2025.03 | 88.3 | 81.43 | 73.33 | |
| ARPL+Backbone=ResNet-50, Pre-training=With Imagenet Pretraining2025.03 | 85.4 | 81.77 | 73.07 | |
| MLSBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=MaxLogit2025.03 | 85.2 | 83.16 | 74.38 | |
| SpHORBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=MaxLogit2025.03 | 82.7 | 87.47 | 76.72 | |
| SpHORBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=PostMax2025.03 | 82.7 | 87.3 | 76.54 | |
| SpHORBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=NNGuide2025.03 | 82.7 | 83.33 | 72.82 | |
| SpHORBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=KNN2025.03 | 82.7 | 84.96 | 75.38 | |
| SpHORBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=MaxLogit, Training Strategy=w/o ROrtho2025.03 | 80.5 | 86.93 | 74.49 | |
| SupCONBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=MaxLogit2025.03 | 78.2 | 88.52 | 72.62 | |
| SupCONBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=PostMax2025.03 | 78.2 | 84.09 | 67.91 | |
| SupCONBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=NNGuide2025.03 | 78.2 | 80.73 | 66.67 | |
| SupCONBackbone=ResNet-50, Pre-training=With Imagenet Pretraining, Scoring Rule=KNN2025.03 | 78.2 | 88.61 | 72.77 | |
| SupCONBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=MaxLogit2025.03 | 77.8 | 87.86 | 71.73 | |
| SupCONBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=PostMax2025.03 | 77.8 | 85.02 | 68.75 | |
| SupCONBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=KNN2025.03 | 77.8 | 88.22 | 72.25 | |
| SupCONBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=NNGuide2025.03 | 77.8 | 81.24 | 66.63 | |
| MLSBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=MaxLogit, Training Strategy=Mixup2025.03 | 56.9 | 70.22 | 45.74 | |
| MLSBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=PostMax, Training Strategy=Mixup2025.03 | 56.9 | 73.1 | 47.1 | |
| MLSBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=NNGuide, Training Strategy=Mixup2025.03 | 56.9 | 59.38 | 38.75 | |
| MLSBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=KNN, Training Strategy=Mixup2025.03 | 56.9 | 52.53 | 31.01 | |
| MLSBackbone=ResNet-50, Pre-training=Without Imagenet Pretraining, Scoring Rule=MaxLogit2025.03 | 53.2 | 68.42 | 43.44 | |
| ARPL+Backbone=ResNet-50, Pre-training=Without Imagenet Pretraining2025.03 | 46.8 | 67.59 | 37.33 |