Open-vocabulary Object Detection on OV-LVIS v1.0 (test)
39.5APrOV-DQUO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OV-DQUOBackbone=ViT-L/14, Training Data=LVIS-base2026.03 | 39.5 | 28.5 | 36 | 33.7 | |
| DeCLIP + NoOVDBackbone=ViT-L/14, Training Data=LVIS-base2026.03 | 39.2 | 38 | 36.9 | 37.7 | |
| CLIPSelf + NoOVDBackbone=ViT-L/14, Training Data=LVIS-base2026.03 | 37.8 | 37.2 | 35.9 | 36.7 | |
| DeCLIP + F-ViTBackbone=ViT-L/14, Training Data=LVIS-base2026.03 | 37.2 | 36.5 | 35.2 | 36 | |
| DeCLIP† + F-ViTBackbone=ViT-L/14, Training Data=LVIS-base, Note=reproduction2026.03 | 36.9 | 36.8 | 35.3 | 36.2 | |
| CLIPSelf† + F-ViTBackbone=ViT-L/14, Training Data=LVIS-base, Note=reproduction2026.03 | 35 | 35.7 | 34.8 | 35.2 | |
| CLIPSelf + F-ViTBackbone=ViT-L/14, Training Data=LVIS-base2026.03 | 34.9 | 35.6 | 34.6 | 35.1 | |
| RO-ViTBackbone=ViT-H/16, Training Data=ALIGN2026.03 | 34.1 | — | — | 35.1 | |
| F-VLMBackbone=RN50x64, Training Data=LVIS-base2026.03 | 32.8 | — | — | 34.9 | |
| RO-ViTBackbone=ViT-L/16, Training Data=ALIGN2026.03 | 32.1 | — | — | 34 | |
| OWL-ViTBackbone=ViT-L/14, Training Data=O365 + VG2026.03 | 31.2 | — | — | 34.6 | |
| F-VLMBackbone=RN50x16, Training Data=LVIS-base2026.03 | 30.4 | — | — | 32.1 | |
| OV-DQUOBackbone=ViT-B/16, Training Data=LVIS-base2026.03 | 29.4 | 23.8 | 27.7 | 26.5 | |
| DeCLIP + NoOVDBackbone=ViT-B/16, Training Data=LVIS-base2026.03 | 29.2 | 31.2 | 23.7 | 27.6 | |
| YOLOEBackbone=YOLOv11-L*, CLIP Initialized=False2026.03 | 29.1 | 36.5 | 35 | 35.2 | |
| SAS-DetBackbone=RN50x4-C42026.03 | 29.1 | 36.8 | 32.4 | 33.5 | |
| CLIPSelf + NoOVDBackbone=ViT-B/16, Training Data=LVIS-base2026.03 | 28.3 | 30.7 | 22.6 | 26.7 | |
| CORA+Backbone=RN50x4, Training Data=LVIS-base + IN-21K2026.03 | 28.1 | — | — | — | |
| RO-ViTBackbone=ViT-B/16, Training Data=ALIGN2026.03 | 28 | — | — | 30.2 | |
| MM-OVODBackbone=RN50*, Training Data=LVIS-base + IN-L, CLIP Initialized=False, Variation=62026.03 | 27.3 | — | — | 33.1 | |
| YOLOEBackbone=YOLOv11-M*, CLIP Initialized=False2026.03 | 26.9 | 34.5 | 32.5 | 33 | |
| DeCLIP + F-ViTBackbone=ViT-B/16, Training Data=LVIS-base2026.03 | 26.8 | 29.8 | 22.4 | 26 | |
| DeCLIP† + F-ViTBackbone=ViT-B/16, Training Data=LVIS-base, Note=reproduction2026.03 | 26.6 | 29.6 | 22.5 | 26 | |
| F-VLMBackbone=RN50x4, Training Data=LVIS-base2026.03 | 26.3 | — | — | 28.5 | |
| DST-DetBackbone=ViT-B/162026.03 | 26.2 | — | — | — | |
| MM-OVODBackbone=RN50*, Training Data=LVIS-base + IN-L, CLIP Initialized=False, Variation=42026.03 | 25.8 | — | — | 32.7 | |
| CLIPSelf† + F-ViTBackbone=ViT-B/16, Training Data=LVIS-base, Note=reproduction2026.03 | 25.4 | 29.3 | 21.8 | 25.4 | |
| CLIPSelf + F-ViTBackbone=ViT-B/16, Training Data=LVIS-base2026.03 | 25.3 | 29.1 | 21.8 | 25.2 | |
| DeticBackbone=RN50*, Training Data=LVIS-base + IN-L, CLIP Initialized=False2026.03 | 24.9 | — | — | 32.4 | |
| MM-OVODBackbone=RN50*, Training Data=LVIS-base + IN-L, CLIP Initialized=False, Variation=52026.03 | 23.8 | — | — | 31.3 | |
| YOLO-WorldBackbone=YOLOv8-L*, Training Data=O365 + GoldG, CLIP Initialized=False2026.03 | 22.9 | 35.4 | 24.9 | 28.7 | |
| DK-DETRBackbone=RN50, Training Data=LVIS-all2026.03 | 22.2 | 40.2 | 32 | 33.5 | |
| CORABackbone=RN50x4, Training Data=LVIS-base2026.03 | 22.2 | — | — | — | |
| LBP2026.03 | 22.2 | 32.4 | 28.8 | 29.1 | |
| RegionCLIPBackbone=RN50x4*, Training Data=CC3M, CLIP Initialized=False2026.03 | 22 | 36.9 | 32.1 | 32.3 | |
| OADPBackbone=ViT-B/32, Training Data=LVIS-all2026.03 | 21.9 | 32 | 28.4 | 28.7 | |
| YOLOEBackbone=YOLOv11-S*, CLIP Initialized=False2026.03 | 21.4 | 29.3 | 26.8 | 27.5 | |
| RKDWTFBackbone=RN50* WTF8x, Training Data=LVIS-base + IN-L, CLIP Initialized=False2026.03 | 21.1 | 29.1 | 25 | 25.9 | |
| SAS-DetBackbone=RN50-C42026.03 | 20.9 | 31.6 | 26.1 | 27.4 | |
| OWL-ViTBackbone=ViT-B/16, Training Data=O365 + VG2026.03 | 20.6 | — | — | 27.2 | |
| YOLO-WorldBackbone=YOLOv8-M*, Training Data=O365 + GoldG, CLIP Initialized=False2026.03 | 19.9 | 32.7 | 22.5 | 26 | |
| MM-OVODBackbone=RN50*, Training Data=LVIS-base, CLIP Initialized=False, Variation=12026.03 | 19.3 | — | — | 30.3 | |
| MM-OVODBackbone=RN50*, Training Data=LVIS-base, CLIP Initialized=False, Variation=32026.03 | 19.3 | — | — | 30.6 | |
| F-VLMBackbone=RN50, Training Data=LVIS-base2026.03 | 18.6 | — | — | 24.2 | |
| MM-OVODBackbone=RN50*, Training Data=LVIS-base, CLIP Initialized=False, Variation=22026.03 | 18.3 | — | — | 29.2 | |
| RKDWTFBackbone=RN50* RKDPIS, Training Data=LVIS-base + IN-L, CLIP Initialized=False2026.03 | 17.3 | 25.5 | 20.9 | 22.1 | |
| RegionCLIPBackbone=RN50*, Training Data=CC3M, CLIP Initialized=False2026.03 | 17.1 | 34 | 27.4 | 28.2 | |
| RKDWTFBackbone=RN50* WTF, Training Data=LVIS-base + IN-L, CLIP Initialized=False2026.03 | 17.1 | 26.7 | 21.4 | 22.8 | |
| YOLO-WorldBackbone=YOLOv8-S*, Training Data=O365 + GoldG, CLIP Initialized=False2026.03 | 13.5 | 26.3 | 16.3 | 19.7 | |
| RKDWTFBackbone=RN50* Base, Training Data=LVIS-base + IN-L, CLIP Initialized=False2026.03 | 12.2 | 26.4 | 19.4 | 20.9 |