averageOfAveragePrecisionAtVariedThresholds(predictions:annotations:confidenceThresholds:)
Calculates average of average precision for all the labels, computed at varied bounding box overlap thresholds. The overlap thresholds range is from [0.05, 0.95] with a stride of 0.05.
Declaration
func averageOfAveragePrecisionAtVariedThresholds<Scalar>(predictions: [[DetectedObject<Label>]], annotations: [ObjectDetectionAnnotation<Label>], confidenceThresholds: [Label : Float] = [:]) -> [Label : Scalar] where Scalar : BinaryFloatingPointParameters
- predictions:
A list of all the predictions from an object detection model. Each element in the list is a list of predictions from one image.
- annotations:
A list of all the annotations. Each element is an
ObjectDetectionAnnotationobject from one image. - confidenceThresholds:
Confidence thresholds for each label. The values will always be between 0.0 and 1.0. If any label does not have a threshold, the
defaultConfidenceThresholdis used for that label. The default value is[:].
Return Value
Average of average precision for all the labels, computed at varied bounding box overlap thresholds.