Pyyan / Object Detection
Finding and boxing things in images and video. The licence matters as much as the accuracy here. · ranked by COCO mAP, and licence realism · verified 13 Aug 2026
| # | Name | COCO mAP | Speed | Licence | Family | Open vocabulary | Status |
|---|---|---|---|---|---|---|---|
| 1 | RF-DETR Roboflow | 60.5 | Real time | Apache 2.0 | Transformer, set prediction | No | current |
| 2 | YOLO26 Ultralytics | ~55 | Very high | AGPL-3.0 | Single stage | No | current |
| 3 | YOLOv12 Ultralytics | ~55 | Very high | AGPL-3.0 | Single stage | No | current |
| 4 | RTMDetOpenMMLab | ~52 | 300+ | MIT | Single stage | No | current |
| 5 | IGrounding DINO IDEA Research | ~52 zero-shot | Moderate | Apache 2.0 | Open vocabulary | Yes | current |
| 6 | Tencent | ~46 zero-shot | High | GPL-3.0 | Open vocabulary | Yes | current |
| 7 | IDINO-X IDEA Research | ~56 | Moderate | Proprietary | Open vocabulary | Yes | current |
| 8 | SAM 2 Meta | Not applicable | Real time | Apache 2.0 | Segmentation | Yes | current |
| 9 | OWLv2 | ~45 zero-shot | Moderate | Apache 2.0 | Open vocabulary | Yes | current |
| 10 | DETR Meta | ~44 | Low | Apache 2.0 | Transformer, set prediction | No | current |
| 11 | Microsoft | ~42 | Low | MIT | Two stage | No | current |
Roboflow
60+
COCO mAP
Ultralytics
43%
faster CPU than YOLO11n