AI, Computer Vision and Machine Learning on Toradex Computer on Modules
Introduction
This article aims to provide an overview of AI, Machine Learning, and Computer Vision on Toradex Computer on Modules. Toradex offers Computer on Modules. The following diagram shows a high-level overview of a typical Computer Vision Application pipeline using Machine Learning. In this article, we will show you the articles that cover each step in the path toward designing an application with AI.
Toradex Computer on Modules tested with AI runtimes
The following Computer-on-Modules (COMs) were already tested with Machine Learning applications:
Computer Vision Development Kits
Built around a Toradex Verdin Computer on Module, Maivin (for 3D Vision) and Raivin (for 4D Radar Sensor Fusion) feature an NXP® i.MX 8M Plus Applications Processor, with an integrated Neural Network Processor with 2.3 TOPS. Learn more.
Cameras
The camera selection is an important task when designing an application for Computer Vision. Toradex computer on modules provides various interfaces to connect a camera, such as Parallel Camera Interface, USB 2.0/3.0, Ethernet, MIPI CSI-2. To help you, we provide an overview of devices tested in our Computer-on-Modules. See the following article for more information:
Overview: Camera on Toradex Computer on ModulesFrame Capture and pre-processing
The process of Frame Capturing consists in collect the image data from the camera driver and make it available as a data array. The pre-process will apply treatments on the image such as resize, padding, filtering, etc, to and prepare the data to the AI runtime effectively process it.
The following articles may help you:
Embedded Linux
Torizon OS
AI / Neural Network Inference
To overcome the challenge of meeting the performance, accuracy, memory footprint requirements of real-world Computer Vision applications, a developer will need to implement AI runtimes optimized for embedded systems. Toradex offers articles with examples implementing some popular AI runtimes with our Computer-on-Modules.
Embedded Linux
Torizon OS
- Torizon Sample: Image Classification with Tensorflow Lite
- Torizon Sample: Real Time Object Detection with Tensorflow Lite
Post-processing and GUI
The optional post-processing step consists of gathering the output data from the AI runtime and create an image to be displayed on a GUI, inserting the data output in overlays with texts, bounding boxes, and graphic information.
Embedded Linux
Torizon OS
Display
In this optional step, the system will make the output data available to the end-user by displaying it in video output. The following article will guide you to the information for Display output with Toradex Computer on Modules:
Additional Resources
Maximizing Edge AI and Security with i.MX 95: Vision, LLMs, and Beyond
Unlock the Power of Embedded Computer Vision: A Deep Dive into Optics, Sensors, and Image Processing
Deep Dive into the i.MX 95 Apps Processor's NPU and Vision Pipeline
Evaluate the real performance of your Edge AI use case without hardware!
Getting Started With Machine Learning Using Toradex and NXP® eIQ
This workshop demonstrates how NXP's eIQ framework can be used to load a model onto the system and conduct inference from a contrived dataset with a Toradex iMX8 SoM running Torizon.