Skip to main content
1

Install RF-DETR

Install the rfdetr package in a Python >= 3.10 environment.
pip install rfdetr
2

Run detection

Load a pre-trained detection model and run inference on an image.
import supervision as sv
from rfdetr import RFDETRMedium
from rfdetr.assets.coco_classes import COCO_CLASSES

model = RFDETRMedium()

detections = model.predict("https://media.roboflow.com/dog.jpg", threshold=0.5)

labels = [f"{COCO_CLASSES[class_id]}" for class_id in detections.class_id]

annotated_image = sv.BoxAnnotator().annotate(detections.data["source_image"], detections)
annotated_image = sv.LabelAnnotator().annotate(annotated_image, detections, labels)
Replace RFDETRMedium with any other class from the model variants reference to use a different size.
3

Run segmentation

Load a pre-trained segmentation model and run inference on an image.
import supervision as sv
from rfdetr import RFDETRSegMedium
from rfdetr.assets.coco_classes import COCO_CLASSES

model = RFDETRSegMedium()

detections = model.predict("https://media.roboflow.com/dog.jpg", threshold=0.5)

labels = [f"{COCO_CLASSES[class_id]}" for class_id in detections.class_id]

annotated_image = sv.MaskAnnotator().annotate(detections.data["source_image"], detections)
annotated_image = sv.LabelAnnotator().annotate(annotated_image, detections, labels)
Replace RFDETRSegMedium with any other segmentation class from the model variants reference to use a different size.
4

Fine-tune on your dataset

Fine-tune a detection or segmentation model on your own labeled dataset.
from rfdetr import RFDETRMedium

model = RFDETRMedium()

model.train(
    dataset_dir="<DATASET_PATH>",
    epochs=100,
    batch_size=4,
    grad_accum_steps=4,
    lr=1e-4,
    output_dir="<OUTPUT_PATH>",
)
The default configuration uses batch_size=4 and grad_accum_steps=4, giving an effective batch size of 16. Adjust both values together to maintain this total across different GPU memory sizes — for example, use batch_size=16, grad_accum_steps=1 on an A100, or batch_size=2, grad_accum_steps=8 on a smaller 12 GB GPU.
RF-DETR automatically detects whether your dataset is in COCO or YOLO format. See the dataset formats guide for details.

​
Next steps

Run detection

Learn all the options for running detection models on images, video, and streams.

Run segmentation

Run segmentation models for pixel-level instance predictions.

Train overview

Full guide to fine-tuning RF-DETR on a custom dataset.

Model variants

All available model sizes with benchmarks and class names.

Build docs developers (and LLMs) love

Get started for freeTalk to us