RF-DETR — Object Detection AI Model
RF-DETR Medium detects the 80 COCO categories in real time, pairing a DINOv2 backbone with a deformable decoder. The pick when frames keep arriving and a slower transformer would fall behind. Boxes plus a text list.
What it does
RF-DETR Medium finds the objects in a photograph and returns each one with a label, a box and a confidence score between 0 and 1, across the 80 everyday COCO categories: people, animals, vehicles, furniture, food, sports gear.
It pairs a DINOv2 backbone with a deformable decoder, which is what gets it to real time. Medium is the balanced size in Roboflow's RF-DETR family - more accurate than the small variants, faster than the large. Out comes an annotated image and a text list, and the threshold trades recall against precision.
When to use it
- Everything in one photograph located at once, in a single pass
- Live video, and other work where the frames keep arriving
- General detection across 80 categories, with no training needed
- A modern transformer detector rather than an older design
- A base to fine-tune on your own categories, as Roboflow intended
Details
- CategoryObject Detection
- Year2025
- LicenseApache-2.0
Compliance & Provenance
- ProviderRoboflow (open) · Specialized
- EU AI Act RiskMinimal Risk
- Art. 50 TransparencyNot applicable
Inputs & Outputs
- ImageInputimage
Source image for object detection
- ImageOutputimage
Detected objects are highlighted with bounding boxes
- TextOutputstring
Detected types of objects with confidence scores
Parameters
- thresholdfloatdefault 0.5range 0–1
Minimum confidence score to keep a detection (higher = fewer, more confident boxes).
Tags
Alternatives in Object Detection
- DETR
General object detection on 91 COCO classes (DETR).
- DETR (Face Detector)
Face detection (DETR ResNet-50 fine-tuned). Fast.
- DETR (License Plate Detector)
License-plate detection (DETR ResNet-50). Fast.
- YOLO-S
Fast object detection, 91 COCO classes (YOLOS-Small).
- YOLO-S (Fashion Item Detection)
Fashion item detection — 46 apparel classes (YOLOS).