Watcher objects are plain object descriptors providing Lampix with the required information to start watching areas based on the behaviors described.

Properties

• name (string): Literal string describing the Python class to instantiate and define the watcher's behavior (e.g 'NeuralNetworkClassifier', 'MovementBasedSegmenter' etc.).

• shape (Object): Plain object with two properties, type and shape that defines the surface area for the watcher.

  • type (string): Defines shape type, and controls what the expected structure of shape.data will be. Accepted values:

    • 'rectangle'

    • 'polygon'

  • data (Object): Actual descriptor object for the contour of the watcher.

    • If shape.type is 'rectangle', the expected data structure is:

      • posX (number): top left X coordinate of rectangle

      • posY (number): top left Y coordinate of rectangle

      • width (number)

      • height (number)

    • If shape.type is 'polygon', a list of points is expected, with each point having the following data structure:

      • x (number): X coordinate of point

      • y (number): Y coordinate of point

• onClassification (Function): Callback used by Lampix to send information about recognized objects as a list with with recognized objects of the following structure:

  • classTag (string): Recognized class of the described object.

  • objectId? (number, Optional): ID representing the object. Useful when determining whether Lampix considers an object as new.

  • outline? (Array, Optional): Object with a property of points that describes the contour of the object as a polygon:

  • metadata? (string, Optional): Watcher specific information

• onLocation (Function): Callback used only by watchers that locate first, then classify. Called with a list of located objects that with a similar structure to the one above, except classTag is not mentioned since it has not been determined at the time of this call.

• params? (Object): Plain object with arbitrary props that can differ from watcher to watcher. See standard watchers​

Notes

The only required property is classTag. Certain watchers that do not support more than one object at a time (e.g NeuralNetworkClassifier), will always provide a list with one recognized object that only has the classTag property. This is by design, to ensure consistency in the data format returned by the standard watchers.

Example

{
  name: 'DepthClassifier',
  shape: {
    type: "rectangle",
    data: {
      posX: 0,
      posY: 0,
      width: window.innerWidth,
      height: window.innerHeight
    }
  },
  params: {
    frames_until_stable: 5
  },
  onClassification: (objects) => draw(objects);
}

Tips

In the example above, you can use shape: lampix.helpers.rectangle(0, 0, window.innerWidth, window.innerHeight) to achieve the same result.