Open-vocabulary detection on LVIS OV (val)
50.1APrLLMDet (Large) + DetRefiner
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LLMDet (Large) + DetRefinerTrained Data=GroundingCap-1M + OV-LVIS2026.05 | 50.1 | 49.3 | 57.1 | 53.2 | |
| LLMDet (Base) + DetRefinerTrained Data=GroundingCap-1M + OV-LVIS2026.05 | 46.1 | 47.3 | 55 | 50.9 | |
| LLMDet (Large)Trained Data=GroundingCap-1M2026.05 | 43.9 | 43.6 | 55.4 | 49.3 | |
| LLMDet (Tiny) + DetRefinerTrained Data=GroundingCap-1M + OV-LVIS2026.05 | 41.9 | 44.5 | 51.9 | 47.9 | |
| MM-Grounding DINO (Base) + DetRefinerTrained Data=O365, GoldG, V3Det + OV-LVIS2026.05 | 41.6 | 43 | 50.4 | 46.4 | |
| MM-Grounding DINO (Tiny) + DetRefinerTrained Data=O365, GoldG, GRIT, V3Det + OV-LVIS2026.05 | 39.4 | 40.7 | 48.1 | 44.2 | |
| LLMDet (Base)Trained Data=GroundingCap-1M2026.05 | 39.2 | 42.1 | 53.1 | 47.2 | |
| LLMDet (Tiny)Trained Data=GroundingCap-1M2026.05 | 36.7 | 37.6 | 49.7 | 43.4 | |
| MM-Grounding DINO (Base)Trained Data=O365, GoldG, V3Det2026.05 | 35.4 | 37.8 | 48.5 | 42.8 | |
| MM-Grounding DINO (Tiny)Trained Data=O365, GoldG, GRIT, V3Det2026.05 | 34.3 | 35.7 | 45.9 | 40.5 | |
| MM-Grounding DINO (Large) + DetRefinerTrained Data=O365V2, OpenImageV6, GoldG + OV-LVIS2026.05 | 34.2 | 38.2 | 43.5 | 40.4 | |
| Grounding DINO (Base) + DetRefinerTrained Data=COCO, O365, GoldG, Cap4M, OpenImages, ODinW35, RefCOCO + OV-LVIS2026.05 | 33.5 | 37.1 | 37 | 36.7 | |
| MM-Grounding DINO (Large)Trained Data=O365V2, OpenImageV6, GoldG2026.05 | 30.4 | 32 | 42.7 | 37 | |
| GLIP (Large) + DetRefinerTrained Data=FourODs, GoldG, CC3M+12M, SBU + OV-LVIS2026.05 | 30.2 | 37.6 | 42.9 | 39.5 | |
| Grounding DINO (Tiny) + DetRefinerTrained Data=O365, GoldG, Cap4M + OV-LVIS2026.05 | 30 | 31.6 | 37.1 | 34.1 | |
| GLIP (Large)Trained Data=FourODs, GoldG, CC3M+12M, SBU2026.05 | 29.1 | 34.7 | 42.3 | 37.9 | |
| Grounding DINO (Base)Trained Data=COCO, O365, GoldG, Cap4M, OpenImages, ODinW35, RefCOCO2026.05 | 27.3 | 31.6 | 36.5 | 33.6 | |
| GLIP (Tiny) + DetRefinerTrained Data=O365, GoldG, CC3M, SBU + OV-LVIS2026.05 | 22.7 | 25.2 | 33.8 | 29.1 | |
| Grounding DINO (Tiny)Trained Data=O365, GoldG, Cap4M2026.05 | 19.9 | 22.6 | 33.1 | 27.4 | |
| GLIP (Tiny)Trained Data=O365, GoldG, CC3M, SBU2026.05 | 18 | 20.5 | 32.3 | 26 |