This report introduces the event "[Jetson Nano Hands-on] Embedded Linux Course for AI Edge Computer Development" held at Macnica Shinagawa office on Thursday, February 6, 2020.
This event was sponsored by our company, Macnica, with the cooperation of Nippon System Development Co., Ltd. (NSK) and NVIDIA LLC (NVIDIA).
hands-on agenda
time |
Lecture title |
speaker |
10:00 - 10:05 |
Opening remarks |
Macnica Kravis Company Takeo Toba |
10:05-10:20 |
Jetson product updates |
Nvidia G.K. Autonomous machine product management Technical Marketing Manager Mr. Yukihiko Tachibana |
10:20-15:55 |
Linux course -Outline of product development flow using AI edge computer Jetson - The underlying architecture of Linux - Software components in Linux distributions - History of Linux - Difference between "original Ubuntu" and "Jetson series Ubuntu" - How to develop each layer
hands-on - Basics of proper use of CPU/GPU - GPU utilization using Jetson Nano - Options for implementing functions in Linux - How to debug |
Japan System Development Co., Ltd. (NSK) Intelligent Technology Department Mr. Masaki Ishihara |
15:55 - 16:00 |
closing remarks |
Macnica Kravis Company Takeo Toba |
12:00-13:00 was a lunch break and we had a light meal.
Why did we hold Jetson Nano Linux Hands-on?
At the beginning, Toba from Macnica gave a greeting. Regarding the background behind planning this event, he said, ``We held this event because we wanted people to know the real challenges of developing AI edge computers.In recent years, high-performance small computers have been released one after another, and services that utilize AI are being developed on edge computers. The number of examples of implementation of AI models on edge computers has increased.Many methods for implementing AI models on edge computers have been published on the Internet, and companies in various industries are working to implement their services in embedded environments. Masu.
However, there are more than a few companies that end up with POC even if they try to implement it in the embedded environment. Although there are multiple factors, one of the major factors we felt while providing support was the lack of basic knowledge for creating embedded products.
Therefore, this time, we would like to give a lecture by NSK, which commercialized AI edge computers with Jetson, and provide an opportunity for customers to acquire the basic knowledge necessary for embedded development. ”
I think it was a moment when the expectations of everyone who participated rose a lot.
Jetson Latest Product Updates
Mr. Tachibana from NVIDIA introduced information that can only be heard here, such as the latest trends at NVIDIA and Jetson Xavier NX, the latest product in the Jetson series.
He introduced some videos that started with the catchphrase "I am ai", and he told us that the BGM used in some of the videos was also composed by AI, which was very interesting. NVIDIA products are used in a variety of situations, such as traffic volume determination, drones and service robots, manufacturing industry inspection equipment and construction machinery for the realization of smart cities, but now they are indispensable as creators. I found out what's going on.
In the latest product introduction, you talked about the latest information on the Jetson series. The high-end model Jetson AGX XAVIER, which has already been released, is equipped with the world's first high-performance processor for autonomous machines. The Jetson Xavier NX is a processor chip that is mounted on a small form factor that is pin-compatible with the Jetson Nano. This is the perfect product for those who have started development with Jetson Nano and want more specs.
Linux course, start!
Mr. Masaki Ishihara of NSK, based on his 20 years of experience in Linux application development, was a lecturer in an embedded Linux course for AI edge computer developers. Although the lecture lasted about three hours, he carefully explained the development flow, technical terms, and the basic knowledge necessary for embedded Linux development.
Start hands-on
For this hands-on training, participants were asked to prepare and bring the following items in advance.
belongings of the day
・Notebook PC (other than Surface, for controlling Jetson Nano)
・ SD card (64GB or more, OS for Jetson Nano installed)
・ AC adapter for Jetson Nano (5 V/4A or more recommended)
・Micro USB-USB cable
Handout
・Jetson Nano
・Software written micro SD card
Finally, you open the Jetson Nano kit, connect it to your own PC, and start practicing.
They proceeded based on the training text while consulting with the instructor and Macnica support members, as well as encouraging each other to speak to their neighbors. We would like to introduce an overview of the training together with questions from participants.
Launch Jetson and connect to PC
How to power Jetson Nano
Jetson Nano is set to micro USB power supply in the initial state. For applications with light software processing, there is no problem with USB power supply, but for heavy processing loads such as automatic driving processing and hands-on this time, we recommend AC adapter power supply. I changed the power supply method to the AC adapter with a jumper pin and started.
Connection with PC
When connecting your PC and Jetson Nano with a USB cable, there were cases where the PC did not recognize the Jetson Nano. At this time, when I restarted Jetson Nano, it started normally.
Execute image processing program (Sobel filter) on CPU/GPU
First, we ran a Sobel filter sample program on the CPU to perform image filtering that emphasizes the borders contained in an image. After that, we changed it to run on the GPU and checked the CPU/GPU usage rate and performance difference at that time. Comparing the measurement time of applying the Sobel filter 1000 times and the processed image, the processing time of the GPU has improved to about 4 times that of the CPU.
The practical training text took about an hour, and everyone completed it and it went smoothly.
Reviewing and supplementing practical training content
Mr. Ishihara of NSK supplemented the know-how that could not be conveyed in a short training program. Among them, in the session "Fundamentals of proper use of CPU/GPU", they enthusiastically explained the programming characteristics unique to GPU using a whiteboard. Many people were enthusiastically taking notes, and he explained the limitations of passing image filtering processing, which GPU is good at, arithmetic processing to CUDA, and how to execute it efficiently.
Need help with embedded Linux development? Contact us
How was the hands-on experience?
Macnica not only acts as a distributor of NVIDIA products, but also collaborates with various partners to help customers solve their problems. If you are starting embedded development using Jetson and have any troubles, please feel free to contact us.