Google's Coral AI Edge Hardware Launches Beta Version

2019-10-23

In March last year, Google released a hardware development kit and accessories called Coral, aimed at supporting the development of edge machine learning models. At that time, it was only beta tested in some regions. However, today, Google announced that Coral will be tested on a "broader" global scale. All Coral products, including the $150 Coral Dev Board, the $74.99 Coral USB Accelerator, and the $2499 5-megapixel camera accessory, can be purchased from electronic retailer Mouser. The company stated that by the end of this year, it will expand distribution to new markets, including but not limited to Taiwan, Australia, New Zealand, India, Thailand, Singapore, Oman, Ghana, and the Philippines. To align with the widespread availability of Coral, Google has made improvements to the Coral website, organizing and arranging documents, tools, recommendations, and pages better. It has also linked them to a new set of examples aimed at providing solutions to common AI problems such as image classification, object detection, pose estimation, and keyword localization. In addition, Google has publicly stated that they will soon release a new version of the Mendel operating system and update the system to the latest version of Debian (Buster). Currently, they are working on updating the Edge TPU compiler to improve the model development workflow. Vikram Tank, the product manager of Coral, wrote: "Over the past six months, the company has received a lot of feedback, and based on this feedback, we have improved our platform. Coral has become the new core of local AI applications in various industries, and several of our partners are incorporating Coral into products that require fast ML inference at the edge." For beginners, the Coral Dev Board is a microcomputer with a removable system module and a custom tensor processing unit (TPU) AI chip. Coral USB, on the other hand, is an accelerator, a USB adapter that primarily accelerates machine learning inference on existing Raspberry Pi and Linux systems. The TPU inside the Coral Dev Board can "concurrently execute" deep feedforward neural networks at a rate of 30 frames per second on high-resolution videos, or perform tasks at a rate of over 100 frames per second on single models such as Mobilenet V2. It can also send and receive data through PCIe and USB, and utilize the Google Cloud IoT Edge software stack for data management and processing. According to Google, the TPU uses liquid cooling technology and is about the size of a quarter. It can be inserted into server racks and used internally to power products such as Google Photos, Google Cloud Vision API calls, and Google search results, enabling offline local processing. If you are interested in the Coral beta version, related product series, and plugin system, you can click on the link at the end of the article to visit the Coral official website for more details. Coral official website: [http://Coral.ai](http://Coral.ai)

RECENT AI NEWS

RECENT AI TOOLS