Explanation Quality Evaluation on Oxford-IIIT Pet (test)
37.2Rank Acc (FeatPerm)ResNet-50
Evaluation results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| ResNet-50Backbone=ResNet-50, Pre-training=IN-1k2026.05 | 37.2 | 45.8 | 35.6 | 35.8 | 36.7 | 50.8 | 22.5 | 50 | 50 | 50 | |
| ViT-B/16Backbone=ViT-B/16, Resolution=224, Pre-training=IN-21k2026.05 | 33.8 | 31.3 | 34.8 | 35 | 34.3 | 50 | 15.5 | 50.1 | 50.1 | 49.9 | |
| ResNet-50Backbone=ResNet-50, Pre-training=scratch2026.05 | 32.8 | 37.3 | 33.6 | 33.9 | 33.1 | 50.9 | 18.1 | 50 | 50 | 49.9 | |
| DenseNet-169Backbone=DenseNet-1692026.05 | 32.4 | 40.6 | 35.4 | 35.5 | 35.1 | 50 | 19.1 | 50.3 | 50.3 | 50 | |
| ViT-L/16Backbone=ViT-L/16, Resolution=224, Pre-training=IN-21k2026.05 | 32.3 | 31.5 | 32.5 | 32.8 | 32 | 49.6 | 15.8 | 49.9 | 49.9 | 49.6 | |
| DenseNet-121Backbone=DenseNet-121, Pre-training=CheXpert, Convergence=did not converge within the training budget2026.05 | 31.7 | 35.7 | 32.9 | 32.8 | 32.8 | 50.6 | 17.3 | 49.9 | 50 | 49.9 | |
| DenseNet-201Backbone=DenseNet-2012026.05 | 31.7 | 36.8 | 35.5 | 35.6 | 35.6 | 48.8 | 17.8 | 50.1 | 50.2 | 50 | |
| ResNet-18Backbone=ResNet-182026.05 | 31.3 | 36.8 | 34.5 | 34.7 | 33.7 | 49 | 18.1 | 50.2 | 50.2 | 50 | |
| ResNet-34Backbone=ResNet-342026.05 | 30.9 | 36 | 34.6 | 34.7 | 34.1 | 48.6 | 17.4 | 50.2 | 50.1 | 50 | |
| ResNet-101Backbone=ResNet-1012026.05 | 30.9 | 41.1 | 34.9 | 34.8 | 34.5 | 49.6 | 19.7 | 50.2 | 50.2 | 50 | |
| DenseNet-121Backbone=DenseNet-1212026.05 | 29.3 | 38.3 | 35.4 | 35.7 | 35.2 | 45.3 | 18.4 | 50.3 | 50.4 | 50 |