PyyanThe AI Index

Pyyan / Compare / DETR vs RF-DETR vs YOLOv12

DETR vs RF-DETR vs YOLOv12

3 of 5

Object Detection · verified 13 Aug 2026

×DETRMetacurrent
×RF-DETRRoboflowcurrent
×YOLOv12Ultralyticscurrent
2 slots left
SpecificationDETRRF-DETRYOLOv12
SummaryThe paper that removed anchors and NMS.First real-time detector past 60 mAP on COCO.Attention-centric YOLO, still widely deployed.
COCO mAP~4460.5~55
SpeedLowReal timeVery high
LicenceApache 2.0Apache 2.0AGPL-3.0
FamilyTransformer, set predictionTransformer, set predictionSingle stage
Open vocabularyNoNoNo
Released202020252025
CategoryObject DetectionObject DetectionObject Detection
OfficialMeta ↗Roboflow ↗Ultralytics ↗

Highlighted rows are where these differ.

DETR

  • Reframed detection as set prediction
  • Slow to train, and every transformer detector descends from it

Best for understanding where RF-DETR came from.

Full spec sheet →

RF-DETR

  • Leads the RF100-VL domain-transfer benchmark
  • DINOv2 backbone, markedly better on occluded objects
  • Apache 2.0, which YOLO is not

Best for most custom detection tasks.

Full spec sheet →

YOLOv12

  • Attention added to the classic single-stage design
  • Same licensing consideration as YOLO26

Best for teams already on the YOLO toolchain.

Full spec sheet →