Object Localization on MS-COCO (val)
47.8Top-1 AccuracyCLIP
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CLIPBackbone=ConvNeXt2026.06 | 47.8 | 90.4 | 89.4 | 59.4 | 89.1 | 82 | |
| DINOv2Backbone=ViT-B/142026.06 | 47.7 | 91.6 | 93.5 | 69 | 93.2 | 77.1 | |
| DINOv3Backbone=ViT-B/162026.06 | 47.7 | 91.5 | 92.1 | 77.9 | 91.1 | 75.4 | |
| DINOv3Backbone=ConvNeXt2026.06 | 47.5 | 90 | 90.3 | 66.3 | 81.4 | 76.5 | |
| CLIPBackbone=ResNet-502026.06 | 47.1 | 88.8 | 50.7 | 44.8 | 75.7 | 53.1 | |
| CLIPBackbone=ViT-B/162026.06 | 46.5 | 89 | 74.5 | 61.8 | 88.8 | 73.3 | |
| DINOv1Backbone=ResNet-502026.06 | 45.9 | 86.2 | 87.1 | 59 | 83.8 | 77.4 | |
| DINOv1Backbone=ViT-B/162026.06 | 44.2 | 83.1 | 51.6 | 16.1 | 90.5 | 84.2 | |
| BEiT-3Backbone=ViT-B/162026.06 | 43 | 81.6 | 80.5 | 60.6 | 84.8 | 73.4 | |
| MAEBackbone=ViT-B/162026.06 | 40.2 | 77.5 | 78.4 | 54.7 | 81.8 | 73.8 |