If you’re working on embedded applications, there’s a good chance that machine learning (ML) is on your radar. For many venturing into this world, gathering the appropriate hardware and software resources necessary to begin your development journey is a big challenge. Where edge AI/ML can sometimes feel esoteric and daunting, a good development platform is necessary to get started in the field.

At Alif Semiconductor, we offer our customers our AI/ML AppKit, a comprehensive evaluation kit for developing edge-AI use cases on our Ensemble MCU family. This article will give you the steps you need to start developing AI/ML applications on Alif Semiconductor’s AI/ML AppKit.

Order your AI/ML Appkit

The first step in this process is to get your hands on an AI/ML AppKit. You can order it directly from the Arrow Electronics website here, or reach out to [email protected] if you prefer to purchase it from Alif directly. 

The Ensemble AI/ML AppKit enables quick software development and evaluation for Endpoint Machine Learning use cases. Powered by an Alif Ensemble E7 fusion processor with four compute cores (two Arm Cortex-M55 and two Cortex-A32), and two machine learning accelerator coprocessors (two Arm Ethos-U55 microNPUs) with ample memory and peripherals, this kit is the ideal platform to build Endpoint ML projects that input sensor data and leverage ML in an extremely power-efficient way.  On board is a camera/image sensor for snapshots or video input, plus motion and sound sensors to round out a variety of Endpoint ML use case scenarios. Specifically, it features a 4” color LCD, an integrated camera module configurable for both “selfie” and viewfinder orientations, and includes PDM and I2S microphones, inertial sensors, and more. Several pre-built applications are available for the AI/ML AppKit so you can quickly get started, which we will dive into below.

Register for Documentation Access

Next, you’ll need to register for an account with Alif Semiconductor to gain access to our extensive set of documentation. This documentation is a goldmine of information, covering everything you need to know about the AppKit’s underlying hardware platform and the software tools necessary for developing and deploying ML models effectively. It’s your one-stop resource for all things AI/ML with Alif Semiconductor.

Registration can be completed here, and, if additional assistance is needed, feel free to reach out to customer support at [email protected]. Do be sure that you register from your company email address, as private email addresses may not be approved.

Kit Start Page

Once registered, you’ll want to head over to the AppKit Start Page. This page is the central hub for all things AppKit, including links to the GitHub repository, user guides, pre-built demos, bill-of-materials, and more.

You can find the start page here.

Download the AppKit User Guide

Next, download the AppKit User Guide. This guide includes step-by-step instructions for setting up your AppKit, ranging from target setup requirements to exploring pre-built demos on the device.

You will want to follow the setup instructions on pages 3-11 to download the Alif Security Toolkit and prepare your kit for initial boot-up. The Security Toolkit is necessary for ensuring secure and efficient initialization of your AppKit, allowing you to proceed confidently in your ML development process.

If you encounter any setup issues, submit a support ticket through the technical support page. Be sure to login first, as this support portal is for registered customers only. 

Download Pre-Built Demos

We’ve provided an archive of pre-built demos to help you hit the ground running in your development process. This archive contains pre-built demo images and the files necessary to create a loadable image for the AI/ML AppKit (Gen 2). Specifically, you’ll find two demos in the archive: one for image classification and one for object detection. You can easily load demos for your AI/ML AppKit with these files.

Download the pre-built demo archive here.

Loading Demos to Your AppKit

Once your demos are downloaded, you’ll want to load them onto your AppKit for testing. The AppKit User Guide provides detailed instructions on how to do this, and you can find them on page 12 (Download the AppKit User Guide).

The process consists of opening the command prompt in the directory where the Alif Security Toolkit is installed, selecting the correct COM port for the SEUART using the maintenance -d command, and running the appropriate batch file for the demo you wish to load. This will ensure that the demo is correctly uploaded and ready for testing on your AppKit.

Optimize Your Models

Optimizing your AI/ML models for Alif devices with the Ethos-U55 NPU is crucial in maximizing performance and efficiency. 

Start by visiting the Alif ML Embedded Evaluation Kit repository on GitHub, which contains comprehensive resources and tools for optimizing your models. The User Guide available in the repository provides detailed instructions on utilizing the Ethos U55 NPU effectively. The guide covers aspects such as configuring the NPU, optimizing model architecture, and leveraging Alif’s software tools for enhanced model performance. 

By following these guidelines, you can ensure that your ML models run efficiently on Alif hardware. This will allow you to fully take advantage of the Ethos-U55’s capabilities for accelerated processing and reduced power consumption, ultimately leading to more robust and responsive applications.

Explore the GitHub Repo

Your next step will be to explore Alif Semiconductor’s SDK on GitHub . This repository offers a wealth of resources to enhance your ML development, including an extensive collection of device drivers, OS packs, demo applications, examples, and various repositories.

The device drivers and OS packs provide the necessary software infrastructure to support various peripherals and operating systems. At the same time, the demo applications and examples offer practical insights into the capabilities and use cases of Alif devices. By leveraging these resources, you can accelerate your development process, troubleshoot issues, and better understand the potential applications of Alif’s hardware and software solutions. 

Alif Supports the Process

If you need support building and deploying your own ML models, Alif Semiconductor offers an extensive AI/ML partner network designed to assist you every step of the way. This network connects you with industry experts who can provide guidance, resources, and technical support tailored to your project’s specific needs. You can start by reaching out to Alif, so that they can match you with the correct partner for your needs. Whether you’re just starting with AI/ML or looking to optimize and scale your model, reach out to Alif to ensure you have access to the right expertise and tools. 

Contact Alif Semiconductor today and take the next steps in your ML journey with confidence and support ([email protected]).