Close

Presentation

Always-On modules demonstrate that tinyML is already here!
TimeThursday, December 9th2:30pm - 3:00pm PST
Location3020
Event Type
Special Session (Research Track)
Virtual Programs
Hosted in Virtual Platform
Topics
Embedded Systems
Machine Learning/AI
DescriptiontinyML is broadly defined as machine learning architectures, techniques, tools and approaches capable of performing on-device analytics for a variety of sensing modalities (vision, audio, motion, chemical, etc.) at “mW” (or below) power range targeting predominately battery operated devices. In this talk therefore, I will highlight how tinyML is fueled by data (at the cloud, the edge and the “tiny”), and show how there are massive opportunities in all verticals where machine intelligence meets the physical world. Through the always-on voice and vision processing paradigms such as Qualcomm’s “always-on vision” module, we will highlight how a holistic hardware-algorithm-software co-design ad the extreme edge can be made possible while meeting all tinyML constraints.