Coral
Coral – machine learning without the cloud
Coral provides accelerator hardware for machine learning that runs on the device itself rather than in a data centre.
The Edge TPU is a processor built for one task alone: executing already-trained neural networks. It does not train models – that still happens on a powerful machine – but it runs inference far faster and with far less power than a general-purpose CPU. On a single-board computer this is the difference between a few images per second and a usable live video stream.
Two consequences matter in practice. Image data stays on the device, which resolves many privacy questions before they arise. And the system keeps working without an internet connection, which counts for anything installed in the field.
The M.2 format fits carrier boards and mini PCs with a suitable slot.
At BerryBase you will find Coral accelerators for image recognition, object detection and edge AI projects.