I am Lao Wen, an embedded engineer who loves learning.
Follow me to become even better together!
Since the advent of ChatGPT, the heat of artificial intelligence technology has been pushed to another level, and more and more enterprises are beginning to focus on whether their business can integrate with industry large models, thereby further improving business efficiency.
At the same time, more and more companies believe that artificial intelligence will trigger a technological revolution that will significantly impact every social practitioner.
As an embedded engineer, I usually engage in embedded software programming or embedded hardware design, which seems unrelated to artificial intelligence. However, embedded artificial intelligence has quietly entered our lives.
For example: facial recognition access control systems (computer vision), smart interactive speakers (natural language processing), autonomous driving technology (in-vehicle mobile detection, environmental perception), etc.
To see the popularity of artificial intelligence in the embedded field, one can observe the product line layout of domestic and foreign processor manufacturers. Unlike international high-end processor manufacturers like Intel and NVIDIA, many domestic chip manufacturers are already laying out NPU neural network processor MPU chips.
Examples include Huawei’s HiSilicon Ascend series, Rockchip’s RK series, Allwinner’s V853, Horizon’s X3, and Bitmain’s K510, etc.
Huawei Ascend AI processor, image source: internet
Just last week, I received the first embedded AI development board independently developed by ZDAS, ATK-DLRV1126. This development board is based on Rockchip’s RV1126 mid-to-high-end MPU chip, and I have the opportunity to learn about embedded artificial intelligence and further expand my embedded knowledge.
Let me briefly introduce the hardware resources of the ATK-DLRV1126 development board. The core board uses Rockchip RV1126 as the main CPU, which is a 4-core Cortex-A7 architecture with a maximum frequency of 1.5GHz, equipped with a 2.0TOPS NPU, an ISP image processor, and an integrated H264/H265 hardware codec, etc.
The core board is designed with an 8-layer gold-plated process, equipped with 2GB DDR4 memory, 8GB EMMC storage chip, and uses a standard 260P SODIMM interface to connect to the development board’s baseboard. The 260P SODIMM interface can also be easily integrated into the customer’s PCB to meet various application needs.
Note: For the hardware resources of the function baseboard, please refer to the development board’s supporting documentation “02【ZDAS】ATK-DLRV1126 Development Board Hardware Reference Manual V1.0.pdf”
Now, let’s talk about software applications. According to the official documentation provided by ZDAS, you can basically set up the development environment from scratch, compile u-boot and Linux kernel, build the buildroot file system, and then flash the firmware to run the applications provided by ZDAS.
This series of operations mainly refers to the following documents: “01【ZDAS】ATK-DLRV1126 Quick Experience V1.2.pdf“, “【ZDAS】ATK-DLRV1126 System Development Manual V1.4.pdf“, “【ZDAS】ATK-DLRV1126 AI Example Testing Guide V1.1 (Introduction).pdf“, as shown in the figure below.
Before learning the ATK-DLRV1126 development board, developers should have a certain foundation in embedded Linux development. On this premise, referring to “【ZDAS】ATK-DLRV1126 System Development Manual V1.4.pdf“, you can basically set up the development environment for RV1126 from scratch and perform various firmware compilation, flashing, application testing, etc.
If developers do not have a foundation in embedded Linux development, it is recommended to start with ZDAS’s imx6ull development board. Given the learning difficulty of this development board, I sincerely do not recommend beginners to dive into this board directly, or they may easily be discouraged.
The factory image of the ATK-DLRV1126 development board has already embedded several embedded AI related applications, such as object recognition SSD, license plate recognition, facial detection, human posture detection, and also includes several audio-video related applications, such as video decoding, driving recorder, live555 streaming, etc.
As ZDAS’s first embedded AI development board, unlike the previous large and comprehensive development boards, the ATK-DLRV1126 development board is designed to be compact and exquisite, much like a handheld computer.
This development board’s learning positioning is also quite precise, mainly aimed at students developing embedded AI (machine vision) and audio-video applications. The resources provided by this development board, as well as the learning materials provided by ZDAS, are sufficient for embedded engineers to step into the door of embedded AI development.
Learning notes on the ATK-DLRV1126 development board will be elaborated in subsequent articles, and I welcome students to closely follow the articles of this public account.
Finally, let’s enjoy some real photos of the development board.
-END-
Previous Recommendations: Click the image to jump to read
What should be noted in embedded software reconstruction?
As an embedded engineer, you must not miss this event closely related to embedded systems!
As an embedded engineer, how to write efficient and clear C language programs?
I am Lao Wen, an embedded engineer who loves learning.
Follow me to become even better together!
