Explore a lightweight face detection method designed for the Himax Ultra-Low Power WE-I Plus AI Processor in this 28-minute conference talk from tinyML Asia 2021. Discover how Justin Kao, a Master Student of Electrical Engineering at National Cheng Kung University in Taiwan, addresses the challenges of implementing computer vision solutions in constrained tinyML environments. Learn about the real-time application's architecture, including power mutation, cascade architecture, and system framework. Examine performance metrics, benchmarks, and data class allocation strategies. Gain insights into scaling techniques and potential applications for this face detection method, which balances accuracy and efficiency in memory-constrained, always-on sensing products.
Lightweight Face Detection for Ultra-Low Power AI Processors - tinyML Asia 2021
Overview
Syllabus
Introduction
Target Platform
The Big Picture
The Target
Relative Work
Power Mutation
Cascade Architecture
System Framework
Performance
Highmix benchmark
Data class allocation
Scaling
Summary
Applications
Audience questions
Influence time
Conclusion
Taught by
tinyML
