Toradex AI Vision Starter Kit for NXP i.MX 8QuadMax

Rapid Creation and Deployment of ML Models at the Edge


Machine Learning-assisted computer vision offers the ability to transform industries by using cameras combined with compute at the edge for immediate analytics to drive appropriate action such as being able to detect a manufacturing defect on an assembly line. The biggest challenge to realizing the benefits of such a solution are the complexities involved in not only training and deploying the model, but also having the appropriate hardware available, tuned, and ready to run the models at the edge cost effectively without sending the entire video stream to the cloud for processing.

Contact Toradex »

Version: 1.0
Last updated: 7/2020
Author: AWS 

Contact Toradex to learn more »


The AI Vision Starter Kit from Toradex offers customers value out-of-the-box. The computer vision starter kit which includes the Toradex System on Module (SOM) Apalis i.MX 8, equipped by NXP i.MX 8QuadMax processor and is integrated to several AWS services such as AWS IoT Greengrass and Amazon SageMaker. Together, these companies collaborated to deliver customers an invaluable solution already integrated, making it possible to build AI at the edge proofs-of-concept (PoCs) quickly without giving up on important factors – security and reliability. Then telemetry data is collected and sent to AWS IoT Core which can be used to monitor the system, refine the models, and maintain optimal performance of the entire system. The kit consists of a Toradex Apalis System-on-Module based on NXP i.MX 8QuadMax applications processor together with an Allied Vision camera. The solution is powered on the edge by AWS IoT Greengrass and Amazon SageMaker Neo. The kit comes out-of-the-box with a pre-trained example machine learning model, providing value right out of the box together with a cloud dashboard to receive data and to monitor telemetry data coming from the edge solution. 


The kit simplifies the process of using computer vision to create machine learning models for object detection by pre-integrating with AWS IoT Greengrass and Amazon SageMaker Neo.

  • Off-the-shelf Solution - The hardware is made to work for several years on a 24/7 schedule. The sample software for the demonstration shortens the path to a final solution.
  • Fast time-to-Market - Proof-of-concept can be created by simply replacing the inference model with a custom one trained using Amazon SageMaker.
  • Easy-to-Use - You don't need a hardware expert or a data scientist to create and deploy what would otherwise be a complex solution.
  • Industrial-grade Hardware - Both NXP and Toradex focus on providing solutions that strive in the extreme, dependable environments.


Hardware-optimized Machine Learning Inference at the Edge
Machine Learning inference is a compute intensive task that often does not meet requirements even on cutting-edge computers. Making good use of specialized hardware units, latest technologies and the correct software is paramount for developing a successful solution that meets the performance demands of an embedded system. This is exactly what is delivered by Amazon SageMaker Neo on the Toradex Apalis i.MX8.

Secure, Reliable, and Resilient Cloud Connectivity
When a device is plugged to the internet, security is highly important – both AWS services as well as Toradex hardware address that necessity. In addition, AWS IoT Greengrass also manages connectivity outages transparently, making the Apalis i.MX 8 a truly reliable on-premises extension of the cloud – a premium-grade edge device.