SDK Version: 2.3.3

Building DX-Stream on Orange Pi 5 Plus

This guide provides comprehensive instructions for building and running DX-Stream on the Orange Pi 5 Plus single-board computer with DEEPX DX-M1 NPU for AI accelerator.

The Orange Pi 5 Plus is a powerful ARM-based development board designed for high-performance edge computing. When paired with the DEEPX DX-M1 NPU, it provides significant acceleration for AI and computer vision applications. This guide covers the complete setup process from SD card preparation to running DX-Stream applications.

DX-Stream Building Process

The overall construction and validation process is broken down into four sequential steps.

Step 1: SD Card Setup
Step 2: DX-RT NPU Driver Installation
Step 3: Build DX-Stream
Step 4: Running Demo Applications

Prerequisites

This section outlines the essential hardware and software components required for building and running DX-Stream on the Orange Pi 5 Plus with the DEEPX DX-M1 NPU. Ensuring these requirements is crucial for a successful installation.

Hardware Requirements

Orange Pi 5 Plus development board
DEEPX DX-M1 NPU module (M.2 form factor)
MicroSD card (32GB or larger, Class 10 recommended)
Power supply (5V/4A minimum)
Host Linux machine for SD card preparation

Software Requirements

7-Zip or similar extraction utility
SD card flashing tool or dd command access

Building Process

Step 1: SD Card Setup

Prepare the base Ubuntu OS environment.

Step 1.1 Download Ubuntu Image

Download the official Orange Pi 5 Plus Ubuntu 22.04 image:

Download Link: Orange Pi 5 Plus Ubuntu 22.04 Image
Image Version: Ubuntu 22.04 Jammy Desktop XFCE with Linux kernel 6.1.43
File Size: Approximately 4GB (compressed)

Step 1.2 Extract Image

Extract the downloaded 7z archive:

7z x Orangepi5plus_1.2.0_ubuntu_jammy_desktop_xfce_linux6.1.43.7z

NOTE

If 7-Zip is not installed, install it using:

# Ubuntu/Debian
sudo apt install p7zip-full

# CentOS/RHEL
sudo yum install p7zip

Step 1.3 Flash Image to SD Card

WARNING

Replace /dev/sda with your actual SD card device. Use lsblk or fdisk -l to identify the correct device.

sudo umount /dev/sda

sudo dd if=./Orangepi5plus_1.2.0_ubuntu_jammy_desktop_xfce_linux6.1.43.img of=/dev/sda bs=4M status=progress

sync

Alternative Methods:

Raspberry Pi Imager: Cross-platform GUI tool
Balena Etcher: User-friendly flashing utility
Win32DiskImager: Windows-specific tool

Step 1.4 Initial Boot

(1) Insert the SD card into the Orange Pi 5 Plus
(2) Connect power supply and peripherals
(3) Power on the device

Default Credentials (if applicable):

Username: orangepi
Password: orangepi

Step 2: DX-RT NPU Driver Installation

Install the necessary drivers and runtime for the DEEPX NPU accelerator.

All following steps should be performed on the Orange Pi 5 Plus itself after successful boot.

Step 2.1 Update Linux Kernel Headers

Install the required kernel headers for driver compilation:

cd /opt
sudo apt install ./linux-headers-current-rockchip-rk3588_1.2.0_arm64.deb

NOTE

The kernel headers package should be pre-installed in the Orange Pi image. If missing, contact DEEPX support.

Step 2.2 Configure Git (Optional)

If you encounter SSL certificate issues with GitHub:

git config --global http.sslVerify false

NOTE

This disables SSL verification globally. For production use, consider configuring proper certificates instead.

Step 2.3 Build and Install NPU Linux Driver

cd ~
mkdir -p dxnn && cd dxnn
git clone https://github.com/DEEPX-AI/dx_rt_npu_linux_driver.git
cd dx_rt_npu_linux_driver/modules
./build.sh
sudo ./build.sh -c install

What this does:

Downloads the NPU kernel driver source code
Compiles the driver for the current kernel
Installs the driver modules and creates device nodes

Step 2.4 Build and Install DX-RT Runtime

cd ~/dxnn
git clone https://github.com/DEEPX-AI/dx_rt.git
cd dx_rt
./install.sh --all
./build.sh

Build Options:

--all: Installs all dependencies and development tools
Alternative: ./install.sh --minimal for runtime-only installation

Step 2.5 Verify DX-RT Installation

Test the NPU driver and runtime installation:

dxrt-cli -s

Expected Output:

DXRT v3.0.0
=======================================================
 * Device 0: M1, Accelerator type
---------------------   Version   ---------------------
 * RT Driver version   : v1.7.1
 * PCIe Driver version : v1.4.1
-------------------------------------------------------
 * FW version          : v2.1.4
--------------------- Device Info ---------------------
 * Memory : LPDDR5 6000 MHz, 3.92GiB
 * Board  : M.2, Rev 1.5
 * Chip Offset : 0
 * PCIe   : Gen3 X4 [01:00:00]

NPU 0: voltage 750 mV, clock 1000 MHz, temperature 40'C
NPU 1: voltage 750 mV, clock 1000 MHz, temperature 40'C
NPU 2: voltage 750 mV, clock 1000 MHz, temperature 40'C
dvfs Disabled
=======================================================

Troubleshooting:

If no device is detected, check M.2 module connection
Verify driver installation with lsmod | grep dx
Check system logs with dmesg | grep -i deepx

Step 3: Build DX-Stream

Compile and install the DX-Stream GStreamer framework and plugins.

Step 3.1 Install Build Dependencies

The Orange Pi 5 Plus Ubuntu image includes pre-configured GStreamer packages with Rockchip hardware acceleration support. Install the additional required packages:

sudo apt-get update
sudo apt install -y python3-pip ninja-build
sudo pip install meson
sudo apt-get install -y libeigen3-dev libjson-glib-dev librdkafka-dev libmosquitto-dev libyuv-dev

Package Descriptions:

python3-pip, ninja-build, meson: Build system components
libeigen3-dev: Linear algebra library for computer vision
libjson-glib-dev: JSON parsing library for GLib
librdkafka-dev: Apache Kafka client library
libmosquitto-dev: MQTT broker client library
libyuv-dev: YUV color space conversion library

Step 3.2 Clone and Build DX-Stream

cd ~/dxnn
git clone https://github.com/DEEPX-AI/dx_stream.git
cd dx_stream
./build.sh

Build Process:

Configures Meson build system
Compiles all GStreamer plugins
Builds sample applications and utilities
Installs plugins to system directories

Build Time: Approximately 10-15 minutes on Orange Pi 5 Plus

Step 3.3 Verify Installation

Check if DX-Stream plugins are properly installed:

gst-inspect-1.0 | grep dx

Expected Output:

dxstream:  dxgather: DX Gather
dxstream:  dxinfer: DX Inference
dxstream:  dxmsgbroker: DX Message Broker
dxstream:  dxmsgconv: DX Message Converter
dxstream:  dxosd: DX On-Screen Display
dxstream:  dxpostprocess: DX Post-process
dxstream:  dxpreprocess: DX Pre-process
...

Step 4: Running Demo Applications

Verify the complete system functionality and NPU acceleration.

Step 4.1 Setup Sample Models and Videos

cd ~/dxnn/dx_stream
./setup.sh

Step 4.2 Run Demo Pipeline

./run_demo.sh

Available Demo Options:

Object detection with YOLO models
Face detection and recognition
Pose estimation
Real-time video processing

For detailed demo instructions, refer to the Installation Guide.

NOTE

Hardware Acceleration. The Orange Pi 5 Plus image includes optimized drivers for:

Video Decode/Encode (VPU): Rockchip RK3588 VPU
GPU Acceleration: Mali-G610 MP4
NPU Acceleration (AI): DEEPX DX-M1 (25 TOPS)

For maximum performance, ensure your pipelines utilize these specific hardware accelerators.

Troubleshooting

This section provides guidance for resolving common issues encountered during the DX-Stream installation or runtime.

Common Issues

1. NPU Not Detected

Verify the NPU module is correctly recognized and the driver is loaded.

# Check PCIe connection
lspci | grep -i deepx
# Verify driver loading
dmesg | grep -i dx_rt

2. GStreamer Plugin Not Found

If DX-Stream plugins are not registered, manually refresh the plugin cache.

# Refresh plugin cache
gst-inspect-1.0 --plugin-path=/usr/local/lib/gstreamer-1.0

3. Build Failures

If the build process fails, clean the environment and rebuild.

# Clean and rebuild
./build.sh

4. Performance Issues

Monitor hardware status if performance is slower than expected.

Verify NPU Status
```
dxrt-cli -s
```
Monitor CPU usage
```
htop
```

NOTE

Next Steps. fter successful installation:

Explore the Pipeline Examples
Learn about Writing Custom Applications
Review Element Documentation for advanced usage

Prerequisites​

Hardware Requirements​

Software Requirements​

Building Process​

Step 1: SD Card Setup​

Step 1.1 Download Ubuntu Image​

Step 1.2 Extract Image​

Step 1.3 Flash Image to SD Card​

Step 1.4 Initial Boot​

Step 2: DX-RT NPU Driver Installation​

Step 2.1 Update Linux Kernel Headers​

Step 2.2 Configure Git (Optional)​

Step 2.3 Build and Install NPU Linux Driver​

Step 2.4 Build and Install DX-RT Runtime​

Step 2.5 Verify DX-RT Installation​

Step 3: Build DX-Stream​

Step 3.1 Install Build Dependencies​

Step 3.2 Clone and Build DX-Stream​

Step 3.3 Verify Installation​

Step 4: Running Demo Applications​

Step 4.1 Setup Sample Models and Videos​

Step 4.2 Run Demo Pipeline​

Troubleshooting​

Common Issues​

Prerequisites

Hardware Requirements

Software Requirements

Building Process

Step 1: SD Card Setup

Step 1.1 Download Ubuntu Image

Step 1.2 Extract Image

Step 1.3 Flash Image to SD Card

Step 1.4 Initial Boot

Step 2: DX-RT NPU Driver Installation

Step 2.1 Update Linux Kernel Headers

Step 2.2 Configure Git (Optional)

Step 2.3 Build and Install NPU Linux Driver

Step 2.4 Build and Install DX-RT Runtime

Step 2.5 Verify DX-RT Installation

Step 3: Build DX-Stream

Step 3.1 Install Build Dependencies

Step 3.2 Clone and Build DX-Stream

Step 3.3 Verify Installation

Step 4: Running Demo Applications

Step 4.1 Setup Sample Models and Videos

Step 4.2 Run Demo Pipeline

Troubleshooting

Common Issues