Image Classification on Breeds (test)
72.5Accuracy (1st Worst Subgroup)DIM
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| DIMType=Data-centric, Backbone=ResNet-342024.03 | 72.5 | 82.1 | 84.6 | 91.3 | 91.5 | 89.7 | |
| Jain et al.Type=Data-centric, Backbone=ResNet-342024.03 | 71.6 | 79.5 | 82.9 | 89.3 | 89.9 | 89.7 | |
| Soft-gDROType=Model-centric (labeled by Ours), Backbone=ResNet-342024.03 | 69.4 | 75.3 | 79.7 | 83 | 84.1 | 82.5 | |
| Soft-DIType=Model-centric (labeled by Ours), Backbone=ResNet-342024.03 | 68.2 | 75.3 | 81.4 | 88.3 | 89.6 | 86.7 | |
| DIType=Model-centric (labeled by Ours), Backbone=ResNet-342024.03 | 67.3 | 74.2 | 79.8 | 87.6 | 88.5 | 86 | |
| gDROType=Model-centric (labeled by Ours), Backbone=ResNet-342024.03 | 66.8 | 72.5 | 75.7 | 78.2 | 82.5 | 80.7 | |
| DIType=Model-centric (labeled by Domino [8]), Backbone=ResNet-342024.03 | 66.1 | 72.3 | 77.9 | 84.2 | 85.3 | 85 | |
| DIType=Model-centric (labeled by Jain et al. [13]), Backbone=ResNet-342024.03 | 65.9 | 71.5 | 76.2 | 83.6 | 85.2 | 84.4 | |
| gDROType=Model-centric (labeled by Domino [8]), Backbone=ResNet-342024.03 | 65.2 | 71.3 | 74.2 | 76.5 | 78.2 | 78.3 | |
| JTTType=Model-centric (Unsupervised), Backbone=ResNet-342024.03 | 64.7 | 70 | 73.2 | 75.9 | 79.2 | 80 | |
| gDROType=Model-centric (labeled by Jain et al. [13]), Backbone=ResNet-342024.03 | 63.2 | 69.6 | 72.8 | 74.1 | 77.9 | 76.9 | |
| EIILType=Model-centric (Unsupervised), Backbone=ResNet-342024.03 | 62.8 | 64.5 | 70.6 | 73.2 | 75.3 | 77 | |
| SubYType=Model-centric (Unsupervised), Backbone=ResNet-342024.03 | 60.8 | 62.9 | 67.5 | 70.3 | 73.5 | 74.1 | |
| LfFType=Model-centric (Unsupervised), Backbone=ResNet-342024.03 | 60.5 | 64.1 | 69.5 | 72.1 | 76.9 | 76.9 | |
| ERMType=Baseline, Backbone=ResNet-342024.03 | 52.4 | 58.4 | 64.4 | 69 | 71.4 | 72.7 |