Object Detection on COCO (AP Size Breakdown)
54.9APLLMDet-T
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LLMDet-Tscale=Tiny2026.02 | 54.9 | 40.1 | 58.3 | 68.6 | |
| ObjEmbed-4Bparameters=4B2026.02 | 53 | 35.6 | 59.6 | 72.2 | |
| ObjEmbed-2Bparameters=2B2026.02 | 52.9 | 35.8 | 59.7 | 72.5 | |
| WeDetect-Bscale=Base2026.02 | 52.1 | 34.8 | 57.1 | 69.2 | |
| DINO-R50backbone=ResNet-502026.02 | 51.2 | 35 | 54.3 | 65.3 | |
| WeDetect-Ref-4Bparameters=4B2026.02 | 50 | 34.7 | 57.6 | 69.2 | |
| WeDetect-Ref-2Bparameters=2B2026.02 | 49.9 | 34 | 58 | 68.9 | |
| Grounding-DINO-Tscale=Tiny2026.02 | 47.9 | 33.4 | 51.2 | 62.2 | |
| LMM-Det-7Bparameters=7B2026.02 | 47.5 | 34.7 | 51.8 | 60.3 | |
| GLIP-Tscale=Tiny2026.02 | 46.1 | — | — | — | |
| VLM-FO1-3Bparameters=3B2026.02 | 44 | — | — | — | |
| OWLv2 (L/14)backbone=L/142026.02 | 37.9 | 24.9 | 41.2 | 53.7 | |
| Qwen3-VL-4Bparameters=4B2026.02 | 29.1 | 12.9 | 33.9 | 54.2 | |
| Qwen3-VL-2Bparameters=2B2026.02 | 16.9 | 6.2 | 20.7 | 36.4 |