As director of AI/ML technologies for NXP, I’m very excited to announce
general availability of NXP’s eIQ machine learning (ML) software
development environment. eIQ ML software is available now with inference
engines and libraries leveraged from the tremendous advancements in open
source machine learning technologies. We have deployed and optimized these
technologies, such as CMSIS-NN, TensorFlow, TensorFlow Lite, OpenCV and Arm
NN, for our popular i.MX RT and i.MX applications processors which are easily
accessed through NXP’s development environments for MCUXpresso and
Yocto (Linux) to provide seamless support for your application development.
Furthermore, eIQ software is accompanied by sample applications in object
detection and voice recognition, to provide you with a starting point in their
deployment of machine learning at the edge.
Machine Learning Is on the Rise, but It’s Just the Tip of the
Iceberg
As ML and artificial intelligence (AI) migrate toward the edge, one of the
biggest challenges is deployment on resource-constrained devices, especially
if you’ve been building your ML applications in the cloud. To run your
models directly on edge devices, those models must be optimized and matched to
an inference engine supporting the specific compute engines (that is, CPU, GPU,
DSP, ML accelerator). eIQ ML software solves this challenge, making it easy
for you to integrate complex hardware components and providing the expertise
for machine learning vision and voice applications.
Build your trained models using public or private cloud-based tools, then
simply bring your models into the eIQ software environment to generate the
appropriate inference engine. The eIQ software integration into our MCUXpresso
and Yocto environments takes care of all ML software dependencies, including
the necessary hardware abstraction layers that connect the advanced machine
learning technology to the underlying compute engines—bridging all the
tools needed to bring your ML models to production.
eIQ software allows for ML applications and use cases to have a life of their
own as it opens the door to mass market implementations of ML-enabled edge
devices.
Getting Started: BYOM – Bring Your Own Machine Learning Model
You can easily get started with eIQ ML software by creating and bringing your
model. Then install MCUXpresso SDK or access NXP’s Yocto build
environment, gather what’s needed for ML inferencing within those
packages, build your application then load it onto the hardware to run it.
For more information on NXP eIQ ML software development environment visit:
eIQ ML Software Development Environment