Single board computers are not a new wonder anymore and the market has already experienced so many different devices based on numerous SoC and SOM. So, why should we be excited this time about the launch of a $150 “Coral Dev Board”? Well, there are multiple reasons – but the most important one is, the Coral Dev Board is manufactured by Google – a relatively new player in the consumer market of SBC and i.MX8M dev boards. Also, the dev board along with a USB accelerator stick is built around Google’s Edge TPU, their purpose-built ASIC designed to run machine learning inference at the edge.
So, what is the Edge TPU? The Edge TPU is a small ASIC designed and built by Google that provides high-performance ML inferencing with a low power cost. For example, it can execute state-of-the-art mobile vision models such as MobileNet v2 at 100+ fps, in a power efficient manner. This allows you to add fast ML inferencing to your embedded AI devices in a power-efficient and privacy-preserving way. For more information on this ASIC, check out googles official documentation.
The NXP i.MX8m based Coral Dev Board appears to be an open-spec design. It would join other open-spec i.MX8M SBCs such as the HummingBoard Pulse. Like SolidRun’s Pulse, the Coral is a sandwich-style board with a removable computer module. The big difference from the expanding field of i.MX8M boards are that the Coral SOM module integrates Google’s Edge TPU neural network co-processor.
The Edge TPU is a stripped-down version of Google’s TPU Unit designed to run TensorFlow Lite ML models on Arm Linux based IoT gateways running on boards like the Coral. The gateways connect to Google Cloud services that are optimized with full-strength Cloud TPU chips to work together via Google’s new Cloud IoT Edge framework. Edge TPU enables concurrent execution of multiple AI models per frame on a high-resolution video at 30fps.