Google’s new Coral Development Board and USB Accelerator are now available to the public, bringing do-it-yourself artificial intelligence (AI) and machine learning (ML) to the masses. But what exactly does that even mean?
Google Coral is a local AI platform and complete toolkit which promises you the ability to build efficient, fast, private and offline products benefiting from machine learning and ushering in a new generation of smart devices. First introduced in March 2018, Google Coral is out of beta as of October 2019.
Machine learning has been playing a role in Google’s search for years now, through the RankBrain AI that has become a major signal in determining your page’s position on the search engine results page (SERP). In fact, Coral isn’t even Google’s first step into DIY artificial intelligence. AIY projects offer vision and voice kits to let you experiment with image and voice recognition through neural networks. These kits use the Intel Movidius chip, while the newly available Coral-branded dev board and USB accelerator use the Edge TPU (tensor-processing unit), both of which are application-specific integrated circuits (ASIC) for AI acceleration. Edge TPU significantly outperforms Movidius, and allows users to concurrently execute multiple state-of-the-art AI models in a power efficient way, and integrates with Cloud TPU and Cloud IoT to deliver end-to-end infrastructure for AI-based solutions.
Okay, but what does that mean for you? Coral enables you to create devices that feature object detection, pose estimation, phrase detection, image segmentation and more. As with the Raspberry PI hardware, the opportunities are seemingly endless. Google even provides pre-compiled models to get you started. Models are built in TensorFlow Lite and then compiled to run on Edge TPU, and supported by Debian Linux or Android Things. With a tiny form factor and $150 price point for the development board and $75 for the USB accelerator, Coral products seems destined to play a major role in the development of innovative AI/ML products.
The two aforementioned Coral products, the development board and the USB accelerator, are designed for prototyping devices and building proofs-of-concept. For incorporating AI/ML into existing/legacy systems, Google offers accelerators in three form factors (Half-size Mini PCIe, M.2 A+E and M.2 B+M) as well as a fully-integrated System-on-Module (SoM). Rounding out the offer are a 5 megapixel camera module and an environmental sensor board, providing temperature, light and humidity data for IoT devices. As with the prototyping products, these all feature very approachable price-points and promise to deliver efficiency and scalability for small businesses and startups to established major enterprises.
Coral is already delivering real-world successes and use-cases across multiple industries. In healthcare, Coral has been used to enable hospitals and care centers to respond quickly to falls, prevent bed sores, improve patient care, and reduce costs. Coral was also behind a polyp detection system helping healthcare professionals improve the accuracy of endoscopic procedures. Coral has also been used by municipalities to accurately measure water utility deliveries, reduce waste, and protect equipment from failure. It may seem cliché to repeat that “the possibilities are endless”, but by being able to set AI processing at the edge, even legacy networks can be modernized without the need for processing at each individual node.
Google has embraced the Internet of Things, and Google Coral products with the Edge TPU are part of this company-wide initiative. These products promise to deliver a powerful and affordable way of building and bringing to market smart solutions. Success, ultimately, will depend not just on creating useful devices but by giving them visibility in an increasingly crowded competitive environment. Whether or not you’re in the high-tech industry, Husaria Marketing’s agile approach was born and bred in IT, and we offer the digital marketing expertise and experience you need to succeed.