Coral platform from Google: AI without the cloud
Google has, with Coral, launched a new platform of hardware and software tools that lets you develop neural networks locally and without using the cloud.
Google has, with Coral, launched a new platform of hardware and software tools that lets you develop neural networks locally.
The Coral USB Accelerator contains an Edge TPU chip for use in a local AI. The Coral USB Accelerator has been designed as an easy-to-connect peripheral to make the Edge TPU available to a Raspberry Pi.
With neural networks that run on the device itself, you can quickly and efficiently add AI to a project while maintaining privacy. Developers create and train their neural networks with TensorFlow before they compile and run them on the Edge TPU boards using the supplied software. Once the compiled network has been installed, all calculations are local on the Edge TPU and no data is sent to the cloud. Without the network delays of the cloud, the performance is increased and you keep the data from the users local and under your own control.
The Coral platform also contains a collection of pre-trained and pre-compiled models that are optimised for the Edge TPU. This makes it easy for developers to get started, and the models can also be adapted to suit your own applications.
The new Coral products and AIY-kits will soon be available from the Elektor Shop, among other places. The Coral product line is an interesting addition to the development tool set for the Raspberry Pi. Artificial Intelligence remains one of the most creative domains in the computer world. We therefore cannot wait to see what kind of prototypes will be developed by the community.
The Coral USB Accelerator contains an Edge TPU chip for use in a local AI. The Coral USB Accelerator has been designed as an easy-to-connect peripheral to make the Edge TPU available to a Raspberry Pi.
With neural networks that run on the device itself, you can quickly and efficiently add AI to a project while maintaining privacy. Developers create and train their neural networks with TensorFlow before they compile and run them on the Edge TPU boards using the supplied software. Once the compiled network has been installed, all calculations are local on the Edge TPU and no data is sent to the cloud. Without the network delays of the cloud, the performance is increased and you keep the data from the users local and under your own control.
The Coral platform also contains a collection of pre-trained and pre-compiled models that are optimised for the Edge TPU. This makes it easy for developers to get started, and the models can also be adapted to suit your own applications.
The new Coral products and AIY-kits will soon be available from the Elektor Shop, among other places. The Coral product line is an interesting addition to the development tool set for the Raspberry Pi. Artificial Intelligence remains one of the most creative domains in the computer world. We therefore cannot wait to see what kind of prototypes will be developed by the community.
