Implementation:Roboflow Rf detr Roboflow Inference SDK

Knowledge Sources	Roboflow Inference RF-DETR
Domains	Deployment, Inference
Last Updated	2026-02-08 15:00 GMT

Overview

External tool documentation for running inference on deployed RF-DETR models via the Roboflow Inference SDK.

Description

The inference Python package provides get_model() to load a deployed model by its ID and model.infer() to run detection on images. It handles image encoding, API communication, and result parsing transparently.

Usage

Use after deploying a model to Roboflow to run inference from any Python environment.

Code Reference

Source Location

• External: inference Python package (Roboflow Inference SDK)

Signature

from inference import get_model

model = get_model(model_id="project_id/version")
result = model.infer(image)

Import

from inference import get_model

I/O Contract

Inputs

Name	Type	Required	Description
model_id	str	Yes	Model identifier ("project_id/version")
image	Union[str, np.ndarray, PIL.Image]	Yes	Image path, URL, array, or PIL Image

Outputs

Name	Type	Description
result	Dict	Detection results with bounding boxes, class labels, and confidence scores

Usage Examples

Run Serverless Inference

from inference import get_model

# Load deployed model
model = get_model(model_id="my-detection-project/1")

# Run inference
result = model.infer("image.jpg")

# Access detections
for prediction in result["predictions"]:
    print(f"Class: {prediction['class']}")
    print(f"Confidence: {prediction['confidence']}")
    print(f"Box: {prediction['x']}, {prediction['y']}, {prediction['width']}, {prediction['height']}")

Related Pages

Implements Principle

• Principle:Roboflow_Rf_detr_Serverless_Inference

Requires Environment

• Environment:Roboflow_Rf_detr_Roboflow_Deployment_Credentials