Open-Vocabulary Object Detection on OV-LVIS
40.9AP NovelLVDor (DeCLIP)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| LVDor (DeCLIP)Backbone=ViT-L/14, Modality=F, Train Data=LVIS-base2026.05 | 40.9 | 37.3 | 38.6 | 38.4 | |
| LVDor (DeCLIP)Backbone=ViT-L/14, Modality=T, Train Data=LVIS-base2026.05 | 40.3 | 37.1 | 38.2 | 38.1 | |
| LVDor (CLIPSelf)Backbone=ViT-L/14, Modality=T, Train Data=LVIS-base2026.05 | 39.8 | 36.2 | 38 | 37.5 | |
| LVDor (CLIPSelf)Backbone=ViT-L/14, Modality=F, Train Data=LVIS-base2026.05 | 39.8 | 36.7 | 37.8 | 37.7 | |
| OV-DQUOBackbone=ViT-L/14, Modality=T, Train Data=LVIS-base2026.05 | 39.5 | 36 | 28.5 | 33.7 | |
| LVDor (DeCLIP)Backbone=ViT-L/14, Modality=I, Train Data=LVIS-base2026.05 | 38.4 | 36.9 | 37.9 | 37.5 | |
| LVDor (CLIPSelf)Backbone=ViT-L/14, Modality=I, Train Data=LVIS-base2026.05 | 37.2 | 36.3 | 37.5 | 36.9 | |
| F-ViT (DeCLIP)Backbone=ViT-L/14, Modality=T, Train Data=LVIS-base2026.05 | 37.2 | 35.2 | 36.5 | 36 | |
| F-ViT (CLIPSelf)Backbone=ViT-L/14, Modality=T, Train Data=LVIS-base2026.05 | 34.9 | 34.6 | 35.6 | 35.1 | |
| F-VLM (CLIP)Backbone=RN50x64, Modality=T, Train Data=LVIS-base2026.05 | 32.8 | — | — | 34.9 | |
| OWL-ViTBackbone=ViT-L/14, Modality=T, Train Data=O365 + VG2026.05 | 31.2 | — | — | 34.6 | |
| LVDor (DeCLIP)Backbone=ViT-B/16, Modality=F, Train Data=LVIS-base2026.05 | 30.7 | 28.2 | 32.3 | 30.2 | |
| F-VLM (CLIP)Backbone=RN50x16, Modality=T, Train Data=LVIS-base2026.05 | 30.4 | — | — | 32.1 | |
| LVDor (DeCLIP)Backbone=ViT-B/16, Modality=T, Train Data=LVIS-base2026.05 | 30.4 | 25.3 | 31.9 | 28.8 | |
| RTGenSupervision=RTGen, Backbone=SwinB (88M), Strict OV=true2024.05 | 30.2 | 39.9 | 41.3 | 38.8 | |
| LVDor (CLIPSelf)Backbone=ViT-B/16, Modality=F, Train Data=LVIS-base2026.05 | 29.8 | 26.8 | 31.5 | 29.2 | |
| LVDor (CLIPSelf)Backbone=ViT-B/16, Modality=T, Train Data=LVIS-base2026.05 | 29.6 | 24.6 | 31.4 | 28.1 | |
| CoDetSupervision=Caption, Backbone=SwinB (88M), Strict OV=true2024.05 | 29.4 | 39.5 | 43 | 39.2 | |
| OV-DQUOBackbone=ViT-B/16, Modality=T, Train Data=LVIS-base2026.05 | 29.4 | 27.7 | 23.8 | 26.5 | |
| YOLOEBackbone=YOLOv11-L*, Modality=T, Train Data=O365 + GoldG2026.05 | 29.1 | 35 | 36.5 | 35.2 | |
| SAS-DetBackbone=RN50x4-C4, Modality=T, Train Data=LVIS-base2026.05 | 29.1 | 32.4 | 36.8 | 33.5 | |
| LVDor (CLIP)Backbone=RN50x4, Modality=F, Train Data=LVIS-base2026.05 | 28.9 | 30.1 | 30.6 | 30.1 | |
| LVDor (CLIP)Backbone=RN50x4, Modality=T, Train Data=LVIS-base2026.05 | 28.8 | 29.7 | 30.4 | 29.8 | |
| CORA+Modality=T, Train Data=LVIS-base + IN-21K2026.05 | 28.1 | — | — | — | |
| LVDor (CLIP)Backbone=RN50x4, Modality=I, Train Data=LVIS-base2026.05 | 27.9 | 29.5 | 29.9 | 29.4 | |
| MM-OVODBackbone=RN50*, Modality=F, Train Data=LVIS-base + IN-L2026.05 | 27.3 | — | — | 33.1 | |
| LVDor (DeCLIP)Backbone=ViT-B/16, Modality=I, Train Data=LVIS-base2026.05 | 26.9 | 28.1 | 31.4 | 29.2 | |
| F-ViT (DeCLIP)Backbone=ViT-B/16, Modality=T, Train Data=LVIS-base2026.05 | 26.8 | 22.4 | 29.8 | 26 | |
| LVDor (CLIPSelf)Backbone=ViT-B/16, Modality=I, Train Data=LVIS-base2026.05 | 26.4 | 27.2 | 30.7 | 28.4 | |
| VLDetSupervision=Caption, Backbone=SwinB (88M), Strict OV=true2024.05 | 26.3 | 39.4 | 41.9 | 38.1 | |
| F-VLMSupervision=CLIP, Backbone=R50x4 (87M), Strict OV=true2024.05 | 26.3 | — | — | 28.5 | |
| F-VLM (CLIP)Backbone=RN50x4, Modality=T, Train Data=LVIS-base2026.05 | 26.3 | — | — | 28.5 | |
| DST-DetBackbone=ViT-B/16, Modality=T, Train Data=LVIS-base2026.05 | 26.2 | — | — | — | |
| MM-OVODBackbone=RN50*, Modality=T, Train Data=LVIS-base + IN-L2026.05 | 25.8 | — | — | 32.7 | |
| F-ViT (CLIPSelf)Backbone=ViT-B/16, Modality=T, Train Data=LVIS-base2026.05 | 25.3 | 21.8 | 29.1 | 25.2 | |
| DeticBackbone=RN50*, Modality=T, Train Data=LVIS-base + IN-L2026.05 | 24.9 | — | — | 32.4 | |
| RTGenSupervision=RTGen, Backbone=RN50-FPN, Strict OV=true2024.05 | 23.9 | 28.5 | 31.8 | 29 | |
| DeticSupervision=Caption, Backbone=SwinB (88M), Strict OV=false2024.05 | 23.9 | 40.2 | 42.8 | 38.4 | |
| MM-OVODBackbone=RN50*, Modality=I, Train Data=LVIS-base + IN-L2026.05 | 23.8 | — | — | 31.3 | |
| CoDetSupervision=Caption, Backbone=RN50-FPN, Strict OV=true2024.05 | 23.4 | 30 | 34.6 | 30.7 | |
| YOLO-WorldBackbone=YOLOv8-L*, Modality=T, Train Data=O365 + GoldG2026.05 | 22.9 | 24.9 | 35.4 | 28.7 | |
| LVDor (CLIP)Backbone=RN50, Modality=F, Train Data=LVIS-base2026.05 | 22.8 | 26.6 | 27.3 | 26.2 | |
| BARONSupervision=CLIP, Backbone=RN50-FPN, Strict OV=true2024.05 | 22.6 | 27.6 | 29.8 | 27.6 | |
| LVDor (CLIP)Backbone=RN50, Modality=T, Train Data=LVIS-base2026.05 | 22.3 | 26.2 | 27.1 | 25.9 | |
| CORABackbone=RN50x4, Modality=T, Train Data=LVIS-base2026.05 | 22.2 | — | — | — | |
| DK-DETRBackbone=RN50, Modality=T, Train Data=LVIS-all2026.05 | 22.2 | 32 | 40.2 | 33.5 | |
| LBPModality=T, Train Data=LVIS-base2026.05 | 22.2 | 28.8 | 32.4 | 29.1 | |
| RegionCLIPSupervision=Caption, Backbone=R50x4 (87M), Strict OV=true2024.05 | 22 | 32.1 | 36.9 | 32.3 | |
| RegionCLIPBackbone=RN50x4*, Modality=T, Train Data=CC3M2026.05 | 22 | 32.1 | 36.9 | 32.3 | |
| OADPBackbone=ViT-B/32, Modality=T, Train Data=LVIS-all2026.05 | 21.9 | 28.4 | 32 | 28.7 | |
| CLIMSupervision=Caption, Backbone=RN50-FPN, Strict OV=true2024.05 | 21.8 | 28.4 | 32 | 28.7 | |
| VLDetSupervision=Caption, Backbone=RN50, Strict OV=true2024.05 | 21.7 | 29.8 | 34.3 | 30.1 | |
| LVDor (CLIP)Backbone=RN50, Modality=I, Train Data=LVIS-base2026.05 | 21.4 | 26.3 | 26.8 | 25.7 | |
| RKDWTFBackbone=RN50* WTF8x, Modality=T, Train Data=LVIS-base + IN-L2026.05 | 21.1 | 25 | 29.1 | 25.9 | |
| SAS-DetBackbone=RN50-C4, Modality=T, Train Data=LVIS-base2026.05 | 20.9 | 26.1 | 31.6 | 27.4 | |
| OWL-ViTBackbone=ViT-B/16, Modality=T, Train Data=O365 + VG2026.05 | 20.6 | — | — | 27.2 | |
| DetProSupervision=CLIP, Backbone=RN50-FPN, Strict OV=true2024.05 | 19.8 | 25.6 | 28.9 | 25.9 | |
| DeticSupervision=Caption, Backbone=RN50, Strict OV=false2024.05 | 19.5 | — | — | 30.9 | |
| MM-OVODBackbone=RN50*, Modality=T, Train Data=LVIS-base2026.05 | 19.3 | — | — | 30.3 | |
| MM-OVODBackbone=RN50*, Modality=F, Train Data=LVIS-base2026.05 | 19.3 | — | — | 30.6 | |
| PromptDetSupervision=Caption, Backbone=RN50-FPN, Strict OV=false2024.05 | 19 | 18.5 | 25.8 | 21.4 | |
| F-VLMSupervision=CLIP, Backbone=RN50-FPN, Strict OV=true2024.05 | 18.6 | — | — | 24.2 | |
| F-VLM (CLIP)Backbone=RN50, Modality=T, Train Data=LVIS-base2026.05 | 18.6 | — | — | 24.2 | |
| MM-OVODBackbone=RN50*, Modality=I, Train Data=LVIS-base2026.05 | 18.3 | — | — | 29.2 | |
| OV-DETRSupervision=Caption, Backbone=RN50-C4, Strict OV=false2024.05 | 17.4 | 25 | 32.5 | 26.6 | |
| RKDWTFBackbone=RN50* RKDPIS, Modality=T, Train Data=LVIS-base + IN-L2026.05 | 17.3 | 20.9 | 25.5 | 22.1 | |
| RegionCLIPSupervision=Caption, Backbone=RN50-C4, Strict OV=true2024.05 | 17.1 | 27.4 | 34 | 28.2 | |
| RegionCLIPBackbone=RN50*, Modality=T, Train Data=CC3M2026.05 | 17.1 | 27.4 | 34 | 28.2 | |
| RKDWTFBackbone=RN50* WTF, Modality=T, Train Data=LVIS-base + IN-L2026.05 | 17.1 | 21.4 | 26.7 | 22.8 | |
| ViLDSupervision=CLIP, Backbone=RN50-FPN, Strict OV=true2024.05 | 16.6 | 24.6 | 30.3 | 25.5 | |
| RKDWTFBackbone=RN50* Base, Modality=T, Train Data=LVIS-base + IN-L2026.05 | 12.2 | 19.4 | 26.4 | 20.9 |