Object Localization on VOC 12 (AP@50%)
72.1AP@50%TokenCut
Evaluation Results
| Method | Links | |
|---|---|---|
| TokenCutBackbone=ViT-S/16, Pre-training=DINO2024.11 | 72.1 | |
| WSCUODBackbone=ViT-S/16, Pre-training=DINO2024.11 | 72.1 | |
| NAVE (N34)Backbone=ResNet50, Pre-training=DINO, Number of clusters (K)=5, Layer settings=N342024.11 | 65 | |
| NAVEBackbone=ViT-S/16, Pre-training=DINO, Number of clusters (K)=5, Output block=N:122024.11 | 64.3 | |
| LOSTBackbone=ViT-S/16, Pre-training=DINO2024.11 | 64 | |
| NAVE (N4)Backbone=ResNet50, Pre-training=DINO, Number of clusters (K)=5, Layer settings=N42024.11 | 62.2 | |
| NAVE (N234)Backbone=ResNet50, Pre-training=DINO, Number of clusters (K)=5, Layer settings=N2342024.11 | 61.7 | |
| CONE-SHAPBackbone=ViT-S/16, Pre-training=DINO, Number of clusters (K)=10, Output block=N:122024.11 | 49.6 | |
| CONE-SHAPBackbone=ResNet50, Pre-training=DINO, Number of clusters (K)=102024.11 | 48.3 | |
| DINO-segBackbone=ViT-S/16, Pre-training=DINO2024.11 | 46.2 | |
| SLICBackbone=ResNet50, Pre-training=DINO2024.11 | 46 | |
| CRAFTBackbone=ResNet50, Pre-training=DINO, Number of clusters (K)=102024.11 | 26.6 |