Close

Presentation

A Domain-Specific System-On-Chip Design for Energy Efficient Wearable Edge AI Applications
TimeWednesday, July 13th6pm - 7pm PDT
LocationLevel 2 Lobby
Event Type
Networking Reception
Work-in-Progress Poster
DescriptionWearable applications collect a significant amount of streaming sensor data. Transmitting the raw data to the cloud wastes the narrow energy budget. Edge AI platforms must balance two competing requirements: (1) Maximizing the energy efficiency using targeted hardware accelerators; and (2) providing developers versatility using general purpose cores for arbitrary applications. Towards this goal, we present an open-source domain-specific programmable system-on-chip that combines a RISC-V core with a meticulously determined set of accelerators targeting wearable applications. Thorough experimental evaluations on an FPGA prototype show 3.6--7.3x higher energy-efficiency and faster execution time than SW implementations of six applications with real-life use-cases.