With the Fathom, Movidius aims to take the advances that have been made in machine learning and apply them to native devices. It envisions the Fathom being used with devices like drones, enabling them to take steps like landing safely on their own, or with surveillance cameras, to apply scene intelligence. Taking these functions off the cloud allows for more privacy and faster decision making, Movidius noted. As Scott Stein wrote on CNET, "Think of it as a sort of mini neural network on the go."
Movidius has other plans in the works for bringing neural networking off the cloud: Back in January, the company inked a deal with Google to use its chips in Android devices, giving the next generation of smartphones more machine-intelligence processing to contextualize images and audio.
The Fathom is compatible with existing deep learning frameworks, including Google TensorFlow. Its target price is under $100, CNET reports.
Seagate Innov8: World's first USB-powered desktop hard drive