.. include:: /keyword.rst

============
TAO on Genio
============

`TAO on Genio`_ offers a list of NeuroPilot compatible pretrained ONNX models
from the `NVIDIA TAO Toolkit`_.
MediaTek's NeuroPilot solution converts these pretrained models to TensorFlow Lite format,
ensuring efficient and optimized inference on supported Genio platforms.
It also provides tutorials for downloading the source model from the `NVIDIA NGC Catalog`_ ,
converting it, and deploying it on a Genio board.

For more information, please refer to `TAO on Genio`_.

.. _TAO on Genio: https://mediatek.gitlab.io/genio/doc/tao/ml-mtk-tao.html
.. _NVIDIA TAO Toolkit: https://developer.nvidia.com/tao-toolkit
.. _NVIDIA NGC Catalog: https://catalog.ngc.nvidia.com/models?filters=&orderBy=scoreDESC&query=tao&page=&pageSize=


FAQ
===

.. dropdown:: What is the difference between TAO on Genio and model hub?

    To evaluate performance of contributed models from around the world,
    refer to :doc:`Model Zoo </sw/yocto/ml-guide/ml-model-hub>` to find converted
    TensorFlow Lite (``.tflite``) models using the MediaTek NeuroPilot Converter,
    as well as DLA models generated from the MediaTek ``ncc-tflite`` compiler.


.. dropdown:: How do I integrate TAO models with cameras and video codecs?

    See :doc:`NNStreamer </tools/nnstreamer>` for real-world use cases that
    combines video and camera pipelines with machine learning models