SSDObjectDetectionModel

class SSDObjectDetectionModel : OnnxInferenceModel

Special model class for detection objects on images with built-in preprocessing and post-processing.

It internally uses ONNXModels.ObjectDetection.SSD trained on the COCO dataset.

Note that output class labels do not correspond to ids in COCO annotations. If you want to evaluate this model on the COCO validation/test set, you need to convert class predictions using appropriate mapping.

Since

0.3

Constructors

SSDObjectDetectionModel

fun SSDObjectDetectionModel()

Functions

open override fun close()

Releases the ONNXRuntime - related resources.

copy

open override fun copy(copiedModelName: String?, saveOptimizerState: Boolean, copyWeights: Boolean): TensorFlowInferenceModel

detectObjects

fun detectObjects(imageFile: File, topK: Int = 5): List<DetectedObject>

fun detectObjects(inputData: FloatArray, topK: Int = 5): List<DetectedObject>

Returns the top N detected object for the given image file sorted by the score.

evaluate

fun evaluate(dataset: Dataset, metric: Metrics): Double

predict

open override fun predict(inputData: FloatArray): Int

fun predict(dataset: Dataset): List<Int>

fun predict(inputData: FloatArray, inputTensorName: String, outputTensorName: String): Int

Predicts the class of inputData.

predictRaw

fun predictRaw(inputData: FloatArray): Map<String, Any>

Returns list of multidimensional arrays with data from model outputs.

predictRawWithShapes

fun predictRawWithShapes(inputData: FloatArray): List<Pair<FloatBuffer, LongArray>>

Returns list of pairs from model outputs.

predictSoftly

fun predictSoftly(inputData: FloatArray): FloatArray

Predicts vector of probabilities instead of specific class in predict method.

open override fun predictSoftly(inputData: FloatArray, predictionTensorName: String): FloatArray

reshape

open override fun reshape(vararg dims: Long)

Chain-like setter to set up input shape.

toString

open override fun toString(): String

Properties

inputDataType

lateinit var inputDataType: OnnxJavaType

Data type for input tensor.

inputDimensions

open override val inputDimensions: LongArray

inputShape

lateinit var inputShape: LongArray

Data shape for prediction.

name

var name: String?

outputDataType

lateinit var outputDataType: OnnxJavaType

Data type for output tensor.

outputShape

lateinit var outputShape: LongArray

Data shape for prediction.

SSDObjectDetectionModel

Since

See also

Constructors

Functions

Properties