Model_Export_Deployment.md

graph LR
    Exporter["Exporter"]
    ONNX_Exporter["ONNX Exporter"]
    Generic_Engine_Exporter["Generic Engine Exporter"]
    TensorFlow_SavedModel_Exporter["TensorFlow SavedModel Exporter"]
    CoreML_Exporter["CoreML Exporter"]
    OpenVINO_Exporter["OpenVINO Exporter"]
    TensorFlow_Lite_Exporter["TensorFlow Lite Exporter"]
    Exporter -- "dispatches to" --> ONNX_Exporter
    Exporter -- "dispatches to" --> Generic_Engine_Exporter
    Exporter -- "dispatches to" --> TensorFlow_SavedModel_Exporter
    Exporter -- "dispatches to" --> CoreML_Exporter
    Exporter -- "dispatches to" --> OpenVINO_Exporter
    Exporter -- "dispatches to" --> TensorFlow_Lite_Exporter
    Generic_Engine_Exporter -- "depends on" --> ONNX_Exporter
    TensorFlow_SavedModel_Exporter -- "depends on" --> ONNX_Exporter

Loading

Details

The Model Export & Deployment subsystem is primarily encapsulated within the doclayout_yolo/engine/exporter.py module. This module contains the core logic for converting trained models into various deployment-ready formats.

Exporter

Acts as the central orchestrator for all model export functionalities. It provides a unified interface for initiating the conversion of a trained DocLayout-YOLO model into various target formats. This component embodies the "Model Export" aspect of the subsystem.

Related Classes/Methods:

• Exporter:143-1125

ONNX Exporter

Responsible for converting the DocLayout-YOLO model into the Open Neural Network Exchange (ONNX) format. ONNX serves as a crucial intermediate representation, enabling interoperability across different ML frameworks and hardware.

Related Classes/Methods:

• export_onnx:354-413

Generic Engine Exporter

Converts the model to a highly optimized, generic engine format, likely for high-performance inference on specific hardware or runtime environments. This often involves further optimization beyond the ONNX format.

Related Classes/Methods:

• export_engine:656-725

TensorFlow SavedModel Exporter

Converts the model to TensorFlow's native SavedModel format, which is a universal serialization format for TensorFlow models, suitable for deployment with TensorFlow Serving, TensorFlow Lite, TensorFlow.js, or other TensorFlow environments.

Related Classes/Methods:

• export_saved_model:727-826

CoreML Exporter

Converts the model to Apple's CoreML format, enabling efficient on-device inference on Apple platforms (iOS, macOS, watchOS, tvOS).

Related Classes/Methods:

• export_coreml:583-654

OpenVINO Exporter

Converts the model to the OpenVINO format, optimizing it for deployment on Intel hardware (CPUs, GPUs, VPUs, FPGAs) using the OpenVINO toolkit.

Related Classes/Methods:

• export_openvino:415-498

TensorFlow Lite Exporter

Converts the model to TensorFlow Lite, a lightweight format designed for mobile and embedded devices, enabling on-device machine learning with low latency and small binary size.

Related Classes/Methods:

• export_tflite:844-857

FAQ