Neural Processing Unit-NPU
 | The NXP i.MX95 Technical Guide documentation from RidgeRun is presently being developed. |
NXP i.MX95 is joining the new generation of edge applications were machine learning capabilities and processing acceleration is key. Machine learning covers multiple markets and domains, is capable of solve complex problems ranging from driver monitoring, production line monitoring, object classification and detection to speech-to-intent. However the cost of such flexibility and capacity is high data processing and computations requirement. One of the most effective ways to solve these high requirements on edge, is to integrate a dedicated processing unit built to handle neural networks that offers an improved inference compute performance and power efficiency. For the NXP i.MX95 such unit is the eIQ Neutron NPU.
In this section, you will explore the NPU capabilities of the i.MX95, including its architecture, software stack, and practical usage.
- eIQ Neutron NPU Overview: Provides an overview of the Neutron-S NPU architecture, including its hardware blocks and software integration.
- Machine Learning Software Stack: Describes the supported inference engines and how the eIQ stack is structured on the i.MX95.
- Enabling NPU Software with Yocto: Explains how to integrate the eIQ stack into a Yocto build and enable Machine Learning support.
- Testing ML Demos: Shows how to run sample applications to validate NPU functionality and Machine Learning performance.