Presentation title

Deploying Artificial Intelligence at the Deep Edge

Authors

Valeria Tomaselli

Institution(s)

STMicroelectronics

Presentation type

Presentation of a company involved in the Embedded Systems business

Abstract

Artificial Intelligence is mostly confined into the Cloud, where the available computing, storage and energy resources seem to be unconstrained.

An architecture consisting of dumb sensors that transfer acquired data to the Cloud, for subsequent processing, would quickly show a number of limitations such as centralization of decisions, scalability and high response time. This scenario would be inadequate in view of the Internet of Things (IoT), in which hundreds of billions of sensors will overwhelm the cloud with too high data rate communication over low-speed networks. Furthermore, carbon emissions from large server farms and the privacy of the information transferred should not be underestimated.

In this context, STMicroelectronics is developing solutions to bring Artificial Intelligence closer to sensors inside micro-controllers, enabling distributed intelligence.

This talk will show flows and tools to automatically map neural networks on ST Micro-controllers.

Some application examples will be shown on data from different sensors (video, audio, inertial sensors).

---

Additional material

• Presentation slides: [pdf]

For more details on this presentation please click the button below: