Model Conversion Tutorial

This chapter provides a step-by-step tutorial of:

  • Convert TensorFlowV2 float model to MediaTek proprietary DLA model.

  • Visualize the model and have a grasp on tensor information.

  • Use Neuron Run-time API to perform model evaluation.

  • Pre-process camera frame data and post-process tensor output data for user-defined metadata framework ( GstInference in this case).

../../../../_images/sw_yocto_ml-guide_neuron-sdk_neuron-dev-flowl-dev_flow_overview.png

Neuron Development Flow Overview

Important

Neuron SDK is only available on the Genio 1200 and Genio 700 platforms and requires NDA access.