Last March, Google took the wraps off of Coral, a collection of hardware kits and accessories intended to bolster AI development at the edge. It launched in select regions in beta before graduating to a wider release last October. And today ahead of the 2020 Consumer Electronics Show, Google announced new additions to the Coral family that will become available later this year.
First up is the Coral Accelerator Module, a multi-chip package that sports Google’s custom-designed Edge tensor processing unit (TPU). The module exposes both PCIe and USB interfaces and can easily integrate into custom PCB designs, and the tech giant says it has been working closely with manufacturing partner Murata to ready the module for shipment in Q1 or Q2 2020.
TPUs are application-specific integrated circuits (ASICs) developed specifically for neural network machine learning. The chip inside the Coral Dev Board — the Edge TPU — can execute multiple computer vision models at 30 frames per second or a single model (like MobileNet V2) at over 100fps, thanks in part to Google’s Cloud IoT Edge data management and processing stack.
Edge TPUs aren’t quite like the chips that accelerate algorithms in Google’s datacenters — those TPUs are liquid-cooled and designed to slot into server racks. Edge TPUs measure about a fourth of a penny in size and handle calculations offline to supplement local microcontrollers and sensors. Moreover, they don’t train machine learning models, but instead run inference with a lightweight version of Google’s TensorFlow machine learning framework dubbed TensorFlow Lite.
Today Google also debuted the Coral Dev Board Mini, which provides a smaller form-factor, lower-power, and lower-cost alternative to the Coral Dev Board. The Mini combines the new Coral Accelerator Module with MediaTek’s 8167s system-on-chip to create a board that “excels” at 720p video encoding and decoding and computer vision use cases. It’ll be available in the first half of 2020.
Google says that additionally, it’ll soon offer new flavors of the Coral System-on-Module with 2GB and 4GB of RAM in addition to the original 1GB configuration. Lastly, it says that Asus will soon make available a single-board computer — Tinker Edge T — powered by the Coral System-on-Module and featuring a range of I/O interfaces, multiple camera connectors, programmable LEDs, and color-coded GPIO header.
“As always, we are always looking for ways to improve the platform … Since our release, we’ve been excited by the diverse range of applications already built on Coral across a broad set of industries that range from healthcare to agriculture to smart cities,” wrote Coral Team director Billy Rutledge in a blog post. “More and more industries are beginning to recognize the value of local AI, where the speed of local inference allows considerable savings on bandwidth and cloud compute costs, and keeping data local preserves user privacy.”