With the rapid development and maturation of artificial intelligence (AI), robots have been allowed to perform tasks that were previously only performed by humans. Quality control on the production line of the factory? Want to spot bugs with AI-driven cameras? How to efficiently process large amounts of medical data? Machine learning can help identify potential tumors and help doctors prevent and treat them in a timely manner.
However, such scenarios can only be done quickly and securely enough. If the AI camera takes a few minutes to process the image in a factory environment, it doesn’t make much sense. If patient data needs to be transferred to the cloud, there is a risk that medical data will be compromised. Google’s Coral wants to address these issues.
In an e-mail to themedia The Verge, Coral Product Manager Vikram Tank explains: “Typically, data from the device is sent to large compute instances that are located in a central data center and the machine learning model can run quickly. Coral is Google’s hardware and software component platform that helps you build devices with local AI — hardware acceleration for neural networks… On the edge device. “
Coral combines the necessary AI software and hardware tools to enable enterprises or developers to build local AI products. Google says more and more industries have recognized the value of localization of AI, so localization not only saves bandwidth and cloud computing costs, but also allows data to be stored on the local side to maintain users’ privacy.
In 2019, Coral has launched its Dev Board, USB Accelerator and PCIE Accelerator portfolio, which has been deployed in 36 countries around the world.
At this year’s CES show, Coral introduced a number of products. The first new product is accelerator Module acceleration module, a multi-chip module that includes Edge TPU ASIC and supports PCIe and USB interfaces that can be embedded in custom print circuit designs, created by Google and Japanese electronic parts manufacturer Murata.
The second product is the Dev Board Mini board, which is smaller than the Dev Board and features the MediaTek 8167s SoC for 720P resolution image encoding/decoding and computer vision, with lower power consumption and more affordable power, both expected in 2020 It was officially listed in the first half of the year. Google has also created different versions of Coral System Module (SoM), with new versions of the 2GB and 4GB configurations in addition to the original 1GB LPDDR4 configuration.