Purpose

Detect RF-DETR finds objects belonging to the selected model’s classes. It publishes detections as rows and draws their boxes over the source. A separate field output turns box positions or confidence into image data for other effects.

Use

Place Detect RF-DETR under Analyze in the Image view. Select an RF-DETR ONNX model and the runner configuration for that model size. The supplied Nano configuration is paired with the Nano model.

Once the configuration loads, connect the picture to image_in. Inspect preview, set Threshold, then use detections or field downstream.

Inputs and outputs

With the supplied detector configuration:

• image_in supplies the picture.
• preview draws the selected detection view.
• field publishes box coordinates, masks or another selected field style.
• detections contains box coordinates, scores and class information.
• count gives the number of detections.
• summary gives combined statistics and the primary detection.

The base port table below changes when the configuration loads.

Controls

• ONNX Model and Runner Config must match. Reload Config applies configuration edits.
• Threshold rejects lower-confidence detections. Maxdetections limits the published count.
• Classfilter restricts classes; its default is all.
• Previewstyle selects the drawn view. Fieldstyle selects the data encoded in field.
• Fit selects Squash or Letterbox when preparing the image for the model.

Notes

Detection boxes cover objects, including the background inside each rectangle. Use Segment SAM3 when an effect needs a mask that follows the object’s outline.

Ports

Inputs

• in image
  RGBA

Outputs

• out image
  RGBA

Parameters

Detect RF-DETR

ONNX Model (Modelpath) Str op('detect_rf_detr').par.Modelpath

Default:

"" (Empty String)

Runner Config (Configpath) Str op('detect_rf_detr').par.Configpath

Default:

"" (Empty String)

Reload Config (Reload) Pulse op('detect_rf_detr').par.Reload

Default:

None