otx.algorithms.detection.adapters.openvino.task#

Openvino Task of Detection.

Classes

BaseInferencerWithConverter(configuration, ...)

BaseInferencerWithConverter class in OpenVINO task.

OpenVINODetectionInferencer(hparams, ...[, ...])

Inferencer implementation for OTXDetection using OpenVINO backend.

OpenVINODetectionTask(task_environment)

Task implementation for OTXDetection using OpenVINO backend.

OpenVINOMaskInferencer(hparams, ...[, ...])

Mask Inferencer implementation for OTXDetection using OpenVINO backend.

OpenVINORotatedRectInferencer(hparams, ...)

Rotated Rect Inferencer implementation for OTXDetection using OpenVINO backend.

OpenVINOTileClassifierWrapper(inferencer[, ...])

Wrapper for OpenVINO Tiling.
class otx.algorithms.detection.adapters.openvino.task.BaseInferencerWithConverter(configuration: dict, model: Model, converter: IPredictionToAnnotationConverter)[source]#

Bases: IInferencer

BaseInferencerWithConverter class in OpenVINO task.

await_all() → None[source]#

Await all running infer requests if any.

enqueue_prediction(image: ndarray, id: int, result_handler: Any) → None[source]#

Runs async inference.

forward(image: Dict[str, ndarray]) → Dict[str, ndarray][source]#

Forward function of OpenVINO Detection Inferencer.

get_saliency_map(prediction: Any)[source]#

Saliency map function of OpenVINO Detection Inferencer.

pre_process(image: ndarray) → Tuple[Dict[str, ndarray], Dict[str, Any]][source]#

Pre-process function of OpenVINO Detection Inferencer.

predict(image: ndarray)[source]#

Predict function of OpenVINO Detection Inferencer.

class otx.algorithms.detection.adapters.openvino.task.OpenVINODetectionInferencer(hparams: DetectionConfig, label_schema: LabelSchemaEntity, model_file: str | bytes, weight_file: str | bytes | None = None, device: str = 'CPU', num_requests: int = 1, model_configuration: Dict[str, Any] = {})[source]#

Bases: BaseInferencerWithConverter

Inferencer implementation for OTXDetection using OpenVINO backend.

Initialize for OpenVINODetectionInferencer.

Parameters:

  • hparams – Hyper parameters that the model should use.

  • label_schema – LabelSchemaEntity that was used during model training.

  • model_file – Path OpenVINO IR model definition file.

  • weight_file – Path OpenVINO IR model weights file.

  • device – Device to run inference on, such as CPU, GPU or MYRIAD. Defaults to “CPU”.

  • num_requests – Maximum number of requests that the inferencer can make. Defaults to 1.

class otx.algorithms.detection.adapters.openvino.task.OpenVINODetectionTask(task_environment: TaskEnvironment)[source]#

Bases: IDeploymentTask, IInferenceTask, IEvaluationTask, IOptimizationTask

Task implementation for OTXDetection using OpenVINO backend.

deploy(output_model: ModelEntity) → None[source]#

Deploy function of OpenVINODetectionTask.

evaluate(output_resultset: ResultSetEntity, evaluation_metric: str | None = None)[source]#

Evaluate function of OpenVINODetectionTask.

explain(dataset: DatasetEntity, explain_parameters: ExplainParameters | None = None) → DatasetEntity[source]#

Explain function of OpenVINODetectionTask.

infer(dataset: DatasetEntity, inference_parameters: InferenceParameters | None = None) → DatasetEntity[source]#

Infer function of OpenVINODetectionTask.

load_config() → Dict[source]#

Load configurable parameters from model adapter.

Returns:

config dictionary

Return type:

ADDict

load_inferencer() → OpenVINODetectionInferencer | OpenVINOMaskInferencer | OpenVINORotatedRectInferencer | OpenVINOTileClassifierWrapper[source]#

load_inferencer function of OpenVINO Detection Task.

optimize(optimization_type: OptimizationType, dataset: DatasetEntity, output_model: ModelEntity, optimization_parameters: OptimizationParameters | None = None)[source]#

Optimize function of OpenVINODetectionTask.

property avg_time_per_image: float | None#

Average inference time per image.

property hparams#

Hparams of OpenVINO Detection Task.

class otx.algorithms.detection.adapters.openvino.task.OpenVINOMaskInferencer(hparams: DetectionConfig, label_schema: LabelSchemaEntity, model_file: str | bytes, weight_file: str | bytes | None = None, device: str = 'CPU', num_requests: int = 1, model_configuration: Dict[str, Any] = {})[source]#

Bases: BaseInferencerWithConverter

Mask Inferencer implementation for OTXDetection using OpenVINO backend.

class otx.algorithms.detection.adapters.openvino.task.OpenVINORotatedRectInferencer(hparams: DetectionConfig, label_schema: LabelSchemaEntity, model_file: str | bytes, weight_file: str | bytes | None = None, device: str = 'CPU', num_requests: int = 1)[source]#

Bases: BaseInferencerWithConverter

Rotated Rect Inferencer implementation for OTXDetection using OpenVINO backend.

class otx.algorithms.detection.adapters.openvino.task.OpenVINOTileClassifierWrapper(inferencer: BaseInferencerWithConverter, tile_size: int = 400, overlap: float = 0.5, max_number: int = 100, tile_ir_scale_factor: float = 1.0, tile_classifier_model_file: str | bytes | None = None, tile_classifier_weight_file: str | bytes | None = None, device: str = 'CPU', num_requests: int = 1, mode: str = 'async')[source]#

Bases: BaseInferencerWithConverter

Wrapper for OpenVINO Tiling.

Parameters:

  • inferencer (BaseInferencerWithConverter) – inferencer to wrap

  • tile_size (int) – tile size

  • overlap (float) – overlap ratio between tiles

  • max_number (int) – maximum number of objects per image

  • tile_ir_scale_factor (float, optional) – scale factor for tile size

  • tile_classifier_model_file (Union[str, bytes, None], optional) – tile classifier xml. Defaults to None.

  • tile_classifier_weight_file (Union[str, bytes, None], optional) – til classifier weight bin. Defaults to None.

  • device (str, optional) – device to run inference on, such as CPU, GPU or MYRIAD. Defaults to “CPU”.

  • num_requests (int, optional) – number of request for OpenVINO adapter. Defaults to 1.

  • mode (str, optional) – run inference in sync or async mode. Defaults to “async”.

predict(image: ndarray) → Tuple[AnnotationSceneEntity, Tuple[ndarray, ndarray]][source]#

Run prediction by tiling image to small patches.

Parameters:

image (np.ndarray) – input image

Returns:

AnnotationSceneEntity features: list including feature vector and saliency map

Return type:

detections