Toradex has launched a $490 “Apalis AI Vision Starter Kit” that runs Torizon Linux and an optimized ML stack with AWS IoT Greengrass on its i.MX8 QuadMax based Apalis iMX8 module. The kit includes Toradex’s Ixora Carrier Board and an Allied Vision Alvium 1500 CSI2 camera.

Toradex announced a machine learning (ML) kit developed in collaboration with Amazon Web Services (AWS) and NXP Semiconductors. The Apalis AI Vision Starter Kit includes its i.MX8 QuadMax based Apalis iMX8 module and Ixora Carrier Board, as well as Allied Vision’s Alvium 1500 machine vision camera.

The Apalis AI Embedded Vision Starter Kit functions as a Linux-driven reference implementation for developing cloud-connected computer vision and ML designs for industrial automation, agriculture, medical equipment, and other applications. A development stack driven by Toradex’s Linux-based Torizon OS incorporates Amazon’s AWS IoT Greengrass middleware to manage secure communications with the cloud while also enabling local, offline processing.

Apalis AI Embedded Vision Starter Kit and Apalis iMX8 module
(click images to enlarge)

Designed to ease the learning curve for cloud-connected ML applications, the development platform includes “step-by-step instructions deploy a comprehensive reference software stack with just a few clicks,” says Toradex. The software builds on ease-of-use features in the newbie-focused Torizon distribution, which features OTA updates, Docker support, and security features.

The setup process opens a local, web-based UI and establishes AWS CloudFormation as a dedicated AWS instance. AWS then enables the device as a target for Amazon SageMaker Neo, enabling developers to optimize a ML model for a specific device. The kit also integrates a trained and optimized ML model.

Apalis AI Embedded Vision Starter Kit software architecture and data flow (left) and Ixora Carrier Board
(click images to enlarge)

Documentation is available on the Toradex Developer Center and source code is hosted on GitHub. A free webinar on Feb. 6 offers attendees a chance to win a free kit (see farther below).

The Apalis iMX8 module ships with the high-end i.MX8 QuadMax SoC with 2x 1.6GHz Cortex-A72 and 4x 1.26GHz Cortex-A53 CPU cores. The SoC also provides 2x 266MHz Cortex-M4F cores, 2x Vivante GC7000LiteXS/VX GPUs, and a 640MHz HIFI4 DSP. The 83 x 45mm module is further equipped with 4GB LPDDR4, 16GB eMMC, and dual-band 802.11ac WiFi with Bluetooth 5.0.

The Apalis AI Embedded Vision Starter Kit deploys the Apalis iMX8 via Toradex’s Ixora Carrier Board. The 125 x 90mm board is equipped with HDMI, LVDS, RGB, 4/5-Wire resistive touch, CSI2, S/PDIF in/out, and analog audio interfaces.

The Ixora Carrier Board provides a Gigabit Ethernet port, dual CAN ports, dual RS232 ports, and microSD, mSATA, and mini-PCIe slots. There’s an RTC and a 5V/3.3V power supply with wide-range, 7-27V DC voltage support.

The third key component in the kit is the MIPI-CSI2 connected Alvium 1500 from Allied Vision. Announced a year ago, the Linux-focused industrial camera features V4L2 drivers that support NXP i.MX6/8 and Nvidia Jetson platforms. Toradex did not mention which version is included in the kit, but it may well be the top-of-the-line model with an On Semi AR0521 sensor at 2592 × 1944 pixels at 60fps for up to 5-megapixels.

Alvium 1500 camera with open frame housing (left) and attached to a Linux development board
(click images to enlarge)

The Alvium 1500 is built around an Alvium vision processor that enables “the configuration of image pre-processing functionalities…directly on the Image Signal Processor in the camera,” says Allied Vision. This on-chip processing, which is enabled by an Image Processing Library (IPL), offloads processing tasks from an embedded system’s main processor.

The Apalis AI kit is further equipped with a camera lens and adapter board for the Alvium 1500, a fan, and a 12VDC, 30W supply. There’s also a heatsink for the Apalis iMX8 module and a variety of cables and adapters.

 
Further information

The Apalis AI Embedded Vision Starter Kit is available for a limited time at $490 for single unit purchase. Customers also receive a $50 AWS credit. More information may be found in Toradex’s announcement and product/shopping page.

A webinar called “Jumpstart Cloud-connected Computer Vision and Machine Learning Designs” will showcase the kit on Feb. 6. Registered attendees will earn a chance to win one of two kit giveaways. More information and registration details may be found here.