Object Detection on COCO Novel Base All 2017
58.7AP NovelVLDet-L
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VLDet-LPre-train Data=O365, Backbone=ViT-L2026.01 | 58.7 | 53.9 | 55.2 | |
| VLDet-BPre-train Data=O365, Backbone=ViT-B2026.01 | 54.2 | 50.4 | 51.2 | |
| CCKT-Det++Pre-train Data=CLIP, Backbone=SwinB2026.01 | 46 | — | 46.2 | |
| OV-DQUOPre-train Data=CLIP, Backbone=RN50x42026.01 | 45.6 | — | 48.1 | |
| BARONPre-train Data=CLIP+COCO Cap, Backbone=RN502026.01 | 42.7 | 54.9 | 51 | |
| RegionCLIPPre-train Data=CC3M, Backbone=RN502026.01 | 31.4 | 57.1 | 50.4 | |
| F-VLMPre-train Data=CLIP, Backbone=RN502026.01 | 28 | — | 39.6 | |
| ViLDPre-train Data=CC3M, Backbone=RN502026.01 | 27.6 | 59.5 | 51.3 | |
| OVR-CNNPre-train Data=COCO Caption, Backbone=RN502026.01 | 22.8 | 46 | 39.9 |