[Workshop] tinyML Talks: AIML SoC for Ultra-Low-Power Mobile and IoT devices

Published: by Creative Commons Licence (Last updated: )

tinyML Talks: AIML SoC for Ultra-Low-Power Mobile and IoT devices

Background

AI Development Process

AI Optimization Process

Acceleration

Computation Optimization

Efficient Control Unit: Address Generation HW Unit

Efficient Control Unit

Highest Time Utilization: Pipeline Architecture

Highest Time Utilization

Memory Optimization

Dequantization

Dequantization

Entropy Compression

Entropy Compression

How the Entropy Decoder Works

Pooling on the Fly

Pooling on the Fly

Pooling on the Fly - Concept