This article was originally published on February 2, 2022
How does an inquisitive high school student prototype an edge AI platform and publish his findings in an IEEE paper? With a Kria™ SOM.
Saad Asad, a junior at Evergreen Valley high school in San Jose, CA, sought to explore the trade-offs of mobile edge computing (MEC) and localized processing in IoT devices—but first needed a hardware setup. Specifically, he wanted to measure the performance, latency, and power advantages of AI compute on a multi-camera security system.
The Kria KV260 Vision AI Starter Kit, the development platform for Kria SOMs, turned out to be a perfect fit not only because of its integrated network and vision interfaces, but the fact that it comes with a pre-built smart camera accelerated application, downloadable for free from the Xilinx App Store. The accelerated application is easily implemented on the KV260 Starter Kit without requiring any experience in FPGA hardware design. Saad was quickly able to jump-start his experiment—which included multiple performance and power benchmarks using built-in measurement tools in Yocto Linux. He also demonstrated that Kria SOMs can run at a very cool temperature. In fact, his project was running at 25oC consuming only 6 watts of power.
As it turned out, the case study was accepted and published in the 2021 IEEE International Conference on Imaging Systems and Techniques (IST) proceedings. The Kria SOM development experience is intended to be a self-enabled path for exploration and experimentation, and Saad did just that with his research.
Saad also points out that from a commercial perspective, FPGAs are inherently well suited for IoT given their adaptability to evolving AI and sensor requirements. We agree.
Since its launch in April 2021, the Kria SOM has been a welcome sign for embedded and AI software developers that have sought to adopt the benefits of FPGA and adaptive computing technologies in their production systems, but didn’t know—or want to know—the ins and outs of hardware design. The desire to make FPGAs more accessible has spread to the academic and maker community as well, and the KV260 Starter Kit is proving to be the on-ramp to broader adoption that we envisioned!
The best way to understand how easy it is to design with the Kria SOM is to see it in action. So, if you’re ready to explore the benefits of FPGAs and the Kria SOM, get yourself a starter kit and we’ll show you step-by-step how to implement your own AI smart camera and put vision AI to work for you in minutes.
In the meantime, check out this short interview with Saad to learn more about this cool project, how he got started, and his plans for the future. We met with him remotely of course!
We’re all very interested in learning more about you – tell us about yourself and your high school milestones to date?
Currently I’m a high school junior at Evergreen Valley High in San Jose, and given the nature of Silicon Valley, I’ve been surrounded by tech my whole life. When I was in fourth grade, I first published a basic game on the App Store, which was a huge eye-opener for me. I started getting more into tinkering around with electronics and computers, and in middle school, I built my first PC. Initially, I was just doing it to hopefully get a faster computer than my ancient laptop I was using, but when I sat down and started building, I found myself awed by all the intricate details in the parts. After I was done, I felt a bit disappointed because I just wanted to dig around more. The love that I felt when building the PC and designing the game has translated into a lot of different projects and ideas over the years.
In high school, I’ve been able to work a bit with the Raspberry Pi, and when my family needs some laptop repair, I’m their guy. Outside of school, I’m also heavily involved with FTC robotics, where our local team competes yearly. Aside from tinkering and robotics, I always love taking courses outside of school or learning a new programming language.
In addition to my STEM-related activities, I play competitive squash at the national level. I am currently ranked 8th in California and 78th in the nation in Boy’s Under 19. I love to stay involved with squash, as it allows me to relieve stress from my schoolwork, while also helping me get a great workout.
What got you to explore this topic?
I started getting into the whole world of SoCs after doing some initial hobbyist projects on a Raspberry Pi 4b board. I was able to make some interesting stuff in my own time, like a Network-Attached Storage device for my home. After playing around a bit with the Pi, I was able to get my hands on a programmable SoC from Cyprus, which really opened my eyes. I started small, making some basic programs that allowed me to control an LED light using a companion app on my phone. This was around the time that I started working extensively with Dr. Hussein, and he took my interest to the next level. He initially proposed the whole idea of our platform evaluation, and I was instantly hooked. The way he explained the entire framework of IoT had me interested right from the get-go. We both then started looking at different platforms to run some basic tests on just as a proof of concept, something we could base our later research on. Long story short, we found the Xilinx Kria platform, and decided to give it a go. Despite my rookie-level experience in hardware design, I was able to navigate the platform surprisingly well, which was a huge plus. Overall, doing initial tests on the Xilinx SOM was a great experience, and I’m looking forward to taking a deeper dive soon.
Did you look at other platforms for your research?
Before working on the research project, I was doing some basic projects with Dr. Hussein using a Cypress Programmable SoC kit. After doing a few of these projects, we decided to move forward with our platform evaluation research. For this, I thoroughly looked at many different platforms from Intel, Microchip, and Nvidia. I then found the Xilinx Kria platform and saw that there were a lot of basic project guides available online for it. After looking through a few of these guides, I found them very straightforward, which was perfect for me given my lack of experience in hardware design. Given the nature of this research, we felt it best to go with the Xilinx platform for its ease of use and out-of-the-box readiness. In short, we did research on other platforms, and ultimately chose the Xilinx Kria platform for its ease of use and accompanying step-by-step guides available online.
Have you ever considered designing with traditional FPGAs before?
I have investigated the basics of FPGAs before and am intrigued by the concept overall. Other than working on some projects using Cypress Programmable SoC kit, I have not had the luxury yet to work with FPGAs hands-on, but I definitely would love to look into it. Given the unique nature of FPGAs in terms of hardware customization, it would be ideal for me to pursue in a future project of mine. I love always trying to expand the scope and complexity of the projects that I do in my free time, and certainly working with FPGAs would help me do just that. So yes, I have considered designing with FPGAs before, and I would love to have any opportunity to be able to do so.
