Get Started
Edge Impulse Object Detection
On this page
  1. Commands
  2. Settings
  3. Output Variables
  4. JSON Output

Edge Impulse Object Detection

Detects objects in the camera video with an object detection model trained in Edge Impulse. The model is uploaded in the analyzer setting Model Path (.eim).

Commands

CommandDescription
Has ObjectChecks whether an object is detected: the label typed in Object if set, any object if it is empty.
Count ObjectsWrites the number of objects with that label (or all objects).
Get Object InfoWrites the class name, confidence, center point, and width and height of one object chosen by Select By.
Get All Objects InfoWrites every detected object as JSON.
Is AnomalyChecks whether the model judged the input as an anomaly.
Get Anomaly ScoreWrites the anomaly detection score.

When a Detection Zone is set, only objects inside it are considered.

Settings

SettingDescription
ObjectThe class (label) name to look for, typed directly. It must exactly match a label in the model; leave it empty to target all objects.
Confidence (%)0 to 100, default 50.
Detection ZoneSee Common Items.
Select ByBest Confidence (default), Index, Largest, Closest.
Index (0-based)The position to pick when Select By is Index. Uses the model's detection order; default 0.
Reference PointSee Common Items.

Output Variables

FieldTypeValue
DetectedDigitalTrue when an object is detected.
CountNumberNumber of detected objects.
Class NameTextLabel of the selected object; an empty string when not detected.
Confidence (%)NumberConfidence of the selected object, 0 to 100; -1 when not detected.
Center Point X / Center Point YNumberPixel coordinates of the selected object's center; -1 when not detected.
Width (Pixels) / Height (Pixels)NumberSize of the selected object's box.
Is AnomalyDigitalTrue when judged as an anomaly.
Anomaly ScoreNumberThe anomaly detection score.
Output (JSON Text)TextArray of every detected object.

JSON Output

A flat array with, per object, label, confidence (0 to 100), the box's top-left x and y, width, and height. Unlike the built-in Object Detection there are no nested bbox and center objects, and x and y are not the center.

[
  { "label": "person", "confidence": 87, "x": 100, "y": 50, "width": 200, "height": 350 },
  { "label": "car", "confidence": 75, "x": 400, "y": 200, "width": 150, "height": 100 }
]