Object detection: Open vocabulary
Detect objects named in a text prompt.
All open vocabulary models
| Model | Accuracy (paper under it) | 4090 latency (input under it) | VRAM peak | License |
|---|---|---|---|---|
| Grounding DINO2 sizes: tiny, base | 48.3 box AP48.4 COCO AP, zero-shot paper | 118 ms1199x800 | 2.9 GiB | Commercial OK |
| LLMDet3 sizes: tiny, base, large | 55.3 box AP | 164 ms1199x800 | 3.0 GiB | Commercial OK |
| MM-Grounding-DINO3 sizes: tiny, base, large | 50.3 box AP50.6 COCO mAP, zero-shot paper | 164 ms1199x800 | 3.0 GiB | Commercial OK |