Accelerators for vision AI on edge devices at Embedded Vision Summit, 2023
Every year brings better, faster, cheaper processors for embedded vision. This enables actual productization of vision AI solutions, so we get to use them in our daily lives. Think of the improvements in face detections with masked images for phone security or intruder alerts for home vision systems. Advancements in semiconductor manufacturing have enabled development and productization of visual AI solutions. Today I wanted to share some latest upcoming research on this topic of hardware architecture advancements at The Embedded Vision Summit, which will be held in May, 22-24, 2023, in Santa Clara, California. Please register for this event with discount codes using the promo code SUMMIT23-SOHINI. https://embeddedvisionsummit.com/
There are many talks at the summit on this topic. Three topics caught my eye. The first topic is titled “Introducing the i.MX93: Your “Go-To” Processor for Embedded Vision” by Srikanth Jagannathan of NXP. In this talk, we will learn about NXP’s recent i.MX 93 applications processor family that provides delivers high performance, low power consumption and incredible versatility while being low priced. We will learn about high-performance Arm Cortex-A55 CPUs, a Cortex-M33 for low-power, real-time operation, and an Arm Ethos U-65 NPU that enables high-performance, cost-effective and energy-efficient ML applications. We will learn about the i.MX 93 processors’ rich set of on chip peripherals and learn about the eIQ software toolkit, which allows developers to create complete system-level applications with ease. https://embeddedvisionsummit.com/2023...
The other talk that I find interesting is titled “Using a Neural Processor for Always-Sensing Cameras” by Sharad Chole of Expedera where we learn about Always sensing cameras that are much like their cousins, the always listening google assistant and alexa. These enable automatically locking and unlocking the device based on whether the owner is looking at the screen or within view of the camera. However continuous sensing can be draining to the phone/system battery. The speakers discuss neural processor (NPU) architectures in the image signal path and provide tips for chipmakers, and system architects to successfully evaluate and deploy an NPU in an always-on camera.
https://embeddedvisionsummit.com/2023...
I am also keen to listen to the session titled “Toward the Era of AI Everywhere” by Lokwon Kim of DEEPX. In this talk, we will learn about DEEPX’s vision of enabling edge AI everywhere. We will learn about key innovations that have led DEEPX to deliver extreme performance, with GPU-like AI accuracy and flexibility, at a significantly low price and power consumption levels. This allows scalability and hardware implementations.
https://embeddedvisionsummit.com/2023...