Thanks to the Rasberry Pi, tech enthusiasts quickly got accustomed to single-board computers (SBCs). As the name suggests, they’re complete computers built on single circuit boards. Due to their small size but full functionality, SBCs make tech development accessible to more people who want to learn about it in a hands-on way.
Google recently announced its new SBC, known as the Coral Dev Board, which people can buy for $149. However, instead of targeting people who are new to developing, it’s geared toward people with more advanced skills.
Google says this product is ideal for working on prototypes of projects that perform on-device Machine Learning, such as Internet of Things (IoT) gadgets. It has an Edge TPU co-processor, which is the brand’s microchip designed for running IoT devices that rely on edge computing to work. One of the advantages of that component is that although it’s a piece of hardware, the chip contains artificial intelligence (AI) software and algorithms.
As such, people developing machine learning algorithms could run trained data sets via edge computing. It’s possible to put the data sets directly on a smart device instead of keeping that information in the cloud. Then, devices could use machine learning without needing network connections.
The co-processor is an add-on part of the main board and part of a system-on-module. The co-processor also offers Wi-Fi and Bluetooth, along with onboard memory and graphics processing. The base section of the board is similarly equipped with capabilities to help people with their projects. There’s a microSD slot, USB ports, an audio jack, plus video and camera interfaces.
A third-party blog about the Coral Dev Board notes people need to create their machine learning models in TensorFlow Lite, not TensorFlow. The next step is to compile them using a web compiler so they work with the Edge TPU co-processor. There are some pre-compiled modules available, too.
The Coral Dev Board runs on the Linux operating system, and people who want to train their machine learning algorithms faster than usual can buy an accessory called a USB Accelerator. It brings a process called inference — which machine learning algorithms use to draw conclusions from statistical algorithms during training — to existing systems.
People can use the USB Accelerator on Linux systems, such as the Coral Dev Board or Raspberry Pi. One of the handiest characteristics of the USB Accelerator is that it plugs into a USB-C port. Users can also use the provided USB-C to USB-A converter cable if they don’t have access to the C-style ports, which feature on some of the most recent computers, such as the new Macs. Those interested in ordering the USB Accelerator can buy the product for $74.99.
Users set up the Coral Dev Board with their serial console program of choice. After they download it or launch that application, Google gives step-by-step instructions for getting everything up and running.
One potential downside of the board — if people intend to use it for other things besides their machine learning projects — is that the documentation from Google about how to set up the product warns connecting a monitor and a keyboard to the board could compromise the overall performance.
People can use some SBCs as desktop computers, especially when they have dedicated operating systems. However, this system is more specific and not built for that purpose. But, besides using it to work with machine learning projects, Google says the Dev Board could suit manufacturers that want to evaluate their products while using in-house custom hardware.
Also, although Google discusses these products on in-house pages, clicking the link to purchase either the Coral Dev Board or the USB Accelerator takes people to a partner site called Mouser Electronics. The Google site says the Coral Dev Board ships within a week, but the page on Mouser Electronics indicates it’s sold out and there’s a seven-week lead time from the factory.
It’s worth clarifying that Coral is Google’s AI-specific branch, and it seems the tech giant has big things in mind for the future. More specifically, Coral is a platform for local AI, and Google promises it’s flexible for startups, large businesses and everything in between. There are already several other Coral products available to explore besides the two mentioned here.
It’ll be especially interesting to see what kind of progress the Coral Dev Board makes possible once more people start using it. Even at this earlier stage, though, it’s still a powerful, yet compact, product of sufficient interest to people working on machine learning IoT projects.
By Kayla Matthews
