core.engines package

Submodules

core.engines.tensorflow_engine module

A Tensorflow_engine module.

This module contains the TensorFlowEngine and related classes, which are used to run inference on a TensorFlow model.

Classes:

TFModelConfig: A Class encapsulates TensorFlow specified configuration for the model. TensorFlowPreprocessor: A class for preprocessing input data for TensorFlow models. TensorFlowPostprocessor: A class for postprocessing output data for TensorFlow models. TensorFlowEngine: A concrete class implementing the InferenceEngine for TensorFlow framework.

class core.engines.tensorflow_engine.TFModelConfig(path: str, dtype: str, target: str, device: str = 'CPU')

Bases: object

A Class encapsulates TensorFlow specified configuration for the model.

This class holds configuration data required to run a TensorFlow model.

Variables

  • _path (str) – The path to the model.

  • _dtype (str) – The data type of the model.

  • _device (str) – The device to run the model on.

  • _target (str) – The target task of the model.

  • _output_layer (list) – The output layers of the model.

  • _drawing (dict) – The drawing configuration of the inference task.

property device: str

The device to run the inference on.

Type

str

property drawing: dict

The drawing configuration of the inference task.

Type

dict

property dtype: str

The data type of the model.

Type

str

property output_layer: list

The output layers of the model.

Type

list

property path: str

The model file path.

Type

str

property target: str

The target of the inference task.

Type

str

class core.engines.tensorflow_engine.TensorFlowEngine(config: core.engines.tensorflow_engine.TFModelConfig)

Bases: core.infereng.InferenceEngine

A concrete class implementing the InferenceEngine for TensorFlow framework.

This class implement verify, preprocess, postprocess, _predict methods defined in InferenceEngine abstract base class for TensorFlow framework.

Variables

  • _model_path (str) – The inference model file path.

  • _dtype (str) – The data type of the inference model.

  • _drawing (dict) – The drawing configuration of the inference task.

  • _output_layer (list) – The output layers of the inference model.

  • _input_size (Tuple[int, int]) – The expected input size of the inference model.

  • _model (dict) – The dictionary representing the inference model.

  • _session (tf.compat.v1.Session) – The Tensorflow Session object of the inference model.

  • _preprocessor (Preprocessor) – The Preprocessor object for preprocessing the input data.

  • _postprocessor (Postprocessor) – The Postprocessor object for postprocessing the output data.

property input_size: Tuple[int, int]

The input size of the TensorFlow model.

Type

Tuple[int, int]

postprocess(frame: numpy.ndarray, outputs: dict) → numpy.ndarray

Implement the postprocess method for TensorFlow models.

The method overrides the postprocess method defined in the InferenceEngine abstract base class.

preprocess(frame: numpy.ndarray) → numpy.ndarray

Implement the preprocess method for TensorFlow models.

The method overrides the preprocess method defined in the InferenceEngine abstract base class.

verify() → bool

Implement the verify method for TensorFlow models.

The method overrides the verify method defined in the InferenceEngine abstract base class.

class core.engines.tensorflow_engine.TensorFlowPostprocessor(target: str, drawing: dict)

Bases: core.processors.postprocessor.Postprocessor

A class for postprocessing output data for TensorFlow models.

This class postprocesses the output data by drawing the results on the output frame.

Variables

_postprocessor (Postprocessor) – A Postprocessor object representing the appropriate postprocessor for the output data of TensorFlow models.

postprocess(frame: numpy.ndarray, outputs: dict) → numpy.ndarray

Implement the postprocess method for TensorFlow models.

The method overrides the postprocess method defined in the Postprocessor abstract base class.

class core.engines.tensorflow_engine.TensorFlowPreprocessor(input_size: Tuple[int, int], dtype: str)

Bases: core.processors.preprocessor.Preprocessor

A class for preprocessing input data for TensorFlow models.

This class preprocesses the input data by resizing and converting to the appropriate data type.

Variables

  • _input_size (Tuple[int, int]) – The expected input size of the model.

  • _dtype (str) – The data type of the model.

preprocess(frame: numpy.ndarray) → numpy.ndarray

Implement the preprocess method for TensorFlow models.

The method overrides the preprocess method defined in the Preprocessor abstract base class.

Module contents

The core.engines package.

This package contains the implementations of engines, which are used to run the model inference.