Beginner’s Guide: Quickly Getting Started with the Orange Pi AIpro Development Board

Orange Pi AIpro development board adopts Ascend AI technology route, whether in appearance, performance, or technical service support, it is very excellent, providing 8/20TOPS powerful computing power, covering the mainstream application scenarios of ecological developers, allowing users to practice various innovative scenarios and providing supporting software and hardware. I believe many friends’ development boards have arrived, let’s quickly experience it together.

Orange Pi AIpro supports users to log in through remote login mode or local display mode to log in to the development board inference task. This time, we will introduce and demonstrate using local display mode as an example.

Remote login mode: Use an RJ45 network cable to connect the development board and PC. After starting the development board, users can remotely log in to the AIpro development board via SSH tools on the PC.

Local display mode: Connect the development board directly to the monitor, mouse, and keyboard. After starting the development board, users can directly use the keyboard and mouse to operate on the displayed graphical desktop.

The list of accessories required for local display mode is as follows:

2.1 Download SD Burning Tools and Ubuntu Image

On the PC, go to the Orange Pi official website, hover the mouse over the “Service and Download” menu, click “Download” in the pop-up menu, find “Orange Pi AIpro” on the download page, then click “Orange Pi AIpro” to enter the download detail page, and download the “official tool” and “ubuntu image” shown in the following image.

The official tools and ubuntu images are on Baidu Netdisk, extract the files according to the page prompts, and save the files on the PC in the following directories:

• Burning tool balenaEtcher:

  Official tool>Linux image burning tool-balenaEtcher>Windows_X64 version>balenaEtcher-Setup-1.18.4.exe

• Ubuntu image: Ubuntu>opiaipro_ubuntu22.04_desktop_aarch64_*.img.xz

2.2 Install the Burning ToolbalenaEtcher

On the PC, double-clickbalenaEtcher-Setup-1.18.4.exe and follow the prompts to install.

2.3 Burn the Ubuntu Image to the SD Card

1. Insert the SD card into the card reader, then insert the card reader into the PC.

2. Open the burning tool balenaEtcher, and follow the tool interface prompts to “burn from file” -> “select target disk” -> “start burning!”

(1) Burn from file, select the ubuntu image downloaded on the PC:

(2) Select the target disk, choose the SD card plugged into the PC:

(3) Start burning

2. Connect the development board. The connection for local display mode is as follows:

After powering on the development board, the green LED indicator light is always on, indicating normal startup.

3. Log in to the development board. After plugging in the power supply for the development board, the login interface on the monitor is as follows:

Enter the username HwHiAiUser and the password (default is Mind@123) to log in to the development board, and the main interface is as shown in the following figure. After logging in, please change the default password and keep the new password safe.

At the bottom of the interface, click icon to open the terminal, as shown in the following figure:

In the terminal, you can enter Linux commands, for example, enter the cd command:

1. Example Introduction

In the terminal window, execute the following command to enter the example directory:

cd /opt/opi_test/ResnetPicture

This example is based on the PyTorch framework’s ResNet50 model for classifying *.jpg images and displaying the Top5 confidence classification ID and category name of the image in the terminal.

The logic of the example code is as follows:

(1) Data preprocessing (decoding + scaling), outputting YUV420SP format, 224*224 resolution images;

(2) Color space conversion and normalization, outputting RGB format, 224*224 resolution images;

(3) Model inference, outputting the confidence of each category the image belongs to;

(4) Data post-processing, the Top5 confidence classification ID and category name of the image.

2. Run the Example

(1) Switch to the root user, enter the password, the default password is Mind@123.

su rootpassword:

(2) Run the example.

In the /opt/opi_test/ResnetPicture example directory, execute the following command to run the example:

cd scriptsbash sample_run.sh

After successful execution, the key prompt information displayed on the screen is as follows. The top1-5 in the prompt information indicates the top 5 categories of image confidence, the index indicates the category identifier, and the value indicates the maximum confidence of that category, class indicates the category.

[INFO] top 1: index[162] value[0.905956] class[beagle][INFO] top 2: index[161] value[0.092549] class[bassetbasset hound][INFO] top 3: index[166] value[0.000758] class[Walker houndWalker foxhound][INFO] top 4: index[167] value[0.000559] class[English foxhound][INFO] top 5: index[163] value[0.000076] class[bloodhound sleuthhound]

By running the first example, we have roughly understood the overall process of software and hardware cooperation. If you need to learn more about the capabilities of the Ascend software stack, how to develop & tune inference applications based on the development board on Ascend, and how to obtain more application examples, you can scan the code to refer to the “Orange Pi AIpro Learning Resources One-stop Navigation” to learn more step by step.