The annual Google I/O conference opened at the Coastline Amphitheater in Mountain View on May 7th, local time in the United States. The conference attracted developers from all over the world. There were many highlights. In addition to showing many elements related to the mobile Internet, Android ecology, smart phone App, etc., there were also 3D models and AR effects, Google Lens updates, and the release of web-based Web-based Duplex, the new hardware product mid-range smartphone Pixel 3a was released.
Among them, in the special lecture on May 9th, Google Coral is also quite concerned by developers.
At the TensorFlow Developer Summit in April this year, Google announced a smart hardware platform-Coral, which is used to solve end-to-end solutions for IoT hardware from prototypes to products. Coral provides a complete local AI toolkit that can easily develop your ideas from prototype to production. It contains hardware components, software tools and content to help you create, train, and run neural networks (NNs) locally on your device. The local feature is very suitable for offline situations with limited connections, and these low-power devices take embedded applications into consideration.
It also has the advantage of keeping user data on the device safe and private. Coral uses Edge TPU specially designed for "edge". This small ASIC is designed by Google to provide high-performance ML reasoning capabilities for low-power devices. It is reported that Coral will provide five products at launch, including new product development, expansion of existing designs and expansion of production.
Coral will work with other Google tools such as TensorFlow and TensorFlow Lite, as well as Google Cloud IoT for connected device management. It can eventually expand from startups to large-scale enterprise uses. Coral products, documentation, sample code, data sheets, and other resources are now available.
At the Google I/O conference in May this year, Google used a dedicated session to introduce Coral to all developers in detail again.
This is a new platform for the development of AI applications on devices and demonstrates its machine learning acceleration capabilities through TensorFlow demonstrations. Coral provides tools that can bring private, fast, and efficient neural network acceleration to your device, enabling you to develop the concept of AI applications from prototype to production. You will also learn the technical specifications of Edge TPU hardware and software tools, as well as the application development process.
In addition to introducing Coral's performance and its hardware development products, Google also introduced at the conference, based on Coral hardware (Coral Dev Board), a mobile app-Model Play developed by a domestic team. It is reported that Model Play is an AI model sharing market for global AI developers.
Model Play not only provides a platform for global developers to display and communicate with AI models, but it can also be used with Coral Dev Board with Edge TPU to accelerate ML inference, preview model operation effects in real time on mobile phones, and help AI expand from prototypes to products.
Developers can publish the AI models they have trained, or they can subscribe to and download the models they are interested in to retrain and expand their AI creativity, and realize the idea-prototype-product process. Model Play also presets various commonly used AI models, such as MobileNetV1, InceptionV2, etc., and supports the submission and release of retrainable models to facilitate users to optimize and fine-tune their business data. After using Model Play to connect to Google Coral hardware, you can directly use the built-in algorithm of the Dev Board development board for migration learning and train your own model. In this process, users do not need to write code, just upload pictures to complete the automatic modeling.
Google also called on the publishers at the meeting to use their imaginations to develop more Coral-based creative products to make the Coral ecosystem more perfect and serve humanity and society. At the same time, the Model Play team is also issuing an AI model call-up order to global developers, soliciting deep learning models based on TensorFlow that can be run on the Google Coral Dev Board, and setting up rich AI development hardware, such as Coral USB Accelerator Accelerators, Coral Dev Board development boards, and Android things development boards are rewarded to encourage more developers to participate in activities and share ideas and ideas with tens of millions of AI developers around the world.