.. include:: /keyword.rst

===============
Model Converter
===============


This page describes supported methods for converting YOLOv5s PyTorch models to TFLite format.
It focuses on two primary conversion paths:

* An open-source converter script from the YOLOv5 project.
* The MediaTek MTK Converter tool, provided as part of the NeuroPilot SDK.

.. figure:: /_asset/ai-workflow-overview-step1.png
   :align: center
   :width: 80%

.. raw:: html

   <br><br>


YOLOv5s Model Conversion
========================

.. toctree::
    :hidden:

    Open Source Converter <model_converter/open_source_converter>
    NeuroPilot Converter Tool <model_converter/neuropilot_converter_tool>

Conversion Methods
------------------

1. **Open-Source Converter**

   * Use the ``export.py`` script provided in the official YOLOv5 repository.
     The script uses TensorFlow Lite Converter to create a converter object and export the YOLOv5 model from PyTorch format to TFLite format.
   * Source link: `YOLOv5 GitHub Repository <https://github.com/ultralytics/yolov5>`_

   .. note::

      Open-source converters may change over time.
      This documentation focuses on an end-to-end flow for YOLOv5s only.
      For other models, frameworks, or third-party converters, additional steps may be required.

2. **NeuroPilot Converter Tool**

   * Download the NeuroPilot SDK All-In-One bundle for the target host platform.
   * Install the NeuroPilot Converter tool that matches the host Python version.
   * Use the MTK Converter to convert the YOLOv5 model from PyTorch format to TFLite format.

Next Steps
----------

For detailed, step-by-step instructions, refer to:

* :doc:`Open Source Converter </ai_hub/ai-workflow/model_converter/open_source_converter>` for using the YOLOv5 ``export.py`` script.
* :doc:`NeuroPilot Converter Tool </ai_hub/ai-workflow/model_converter/neuropilot_converter_tool>` for using the MTK Converter.