Overview
Explore a keynote address from tinyML Asia 2022 focusing on innovative approaches for Internet of Video Things (IoVT) applications. Delve into the potential of in-memory computing and dynamic vision sensors to overcome bandwidth and energy limitations in IoT vision sensors. Learn about event-driven dynamic vision sensors and their applications in unattended ground sensors and intelligent transportation systems. Discover how in-memory computing techniques can be utilized to perform basic image processing operations and trigger deep neural networks, further reducing energy consumption in sensor nodes. Gain insights into the novel concept of hybrid in-memory computing, combining multiple types of memory for enhanced efficiency. Throughout the talk, examine technical details, demonstrations, and discussions on energy efficiency in the context of neuromorphic technologies and their impact on IoVT scalability.
Syllabus
Introduction
What are Neuromorphic Technologies
Dynamic Vision Sensors
Memory Computing
Technical Details
Inmemory Computing
Event Based Binary Image
Spatial Sparsity
Connected Component Labelling
Histogram Based Region
Transistors
Demonstration
Hybrid RAM
Energy Efficiency
Conclusion
Sponsors
Taught by
tinyML