Enabling eIQ Core NPU Delegates for i.MX Android Applications

This application note provides detailed instructions on deploying the TensorFlow Lite inference engine and its associated delegates for on-chip machine learning accelerators, specifically the NPU and GPU, on i.MX 8MP and i.MX 95 application processors. The document outlines the primary objective of enabling the use of machine learning hardware accelerators to achieve performance levels comparable to those on NXP Yocto Linux. It includes sections that compare the Neural Networks API (NNAPI) with dedicated delegates, describe the steps to build the Android platform, and explain how to configure the eIQ Core using the Android Native Development Kit (NDK). Additionally, it covers the installation and execution of applications that utilize these hardware accelerators. The document specifies the requirements for following the application note, including necessary hardware and software configurations, and provides guidance on configuring SELinux labels for native libraries, which is essential for enabling applications to run in restrictive modes.