Installing Qt Edge AI from Debian packages

Qt Edge AI is a Technical Preview. This page describes how to install Qt Edge AI from the enterprise Qt Debian package repositories on an NVIDIA Jetson AGX Orin or Jetson Thor device that runs Jetson Linux, which is Debian-based. It also describes how to build the Simple Face Detection example on the device.

Limitations

With Debian packages, Qt Edge AI Technical Preview 1.0 does not provide a way to cross-compile applications on a host. Instead, build your applications on the target device, an NVIDIA Jetson AGX Orin or Jetson Thor. To cross-compile, build Qt and Qt Edge AI yourself.

Prerequisites

To follow this page, you need:

• An NVIDIA Jetson AGX Orin or Jetson Thor device with JetPack 7.x, which runs Jetson Linux r38.x or r39.x, based on Ubuntu 24.04.
• NVIDIA DeepStream 9.1 on the device. For all prerequisites, see Qt Edge AI prerequisites.
• A network connection on the device.
• A Qt Account with a Qt for Device Creation license or an evaluation license. The repository requires the credentials of this account.
• A keyboard and a display, or an SSH connection, to work on the device.

For general information about the Qt Debian packages, see Enterprise Qt Debian Packages. This page describes only the steps that are specific to Qt Edge AI.

Version compatibility

Installing dependencies

Installing NVIDIA DeepStream SDK

Install the DeepStream SDK as instructed in Installation - DeepStream Documentation. Use NVIDIA's Debian package, deepstream-9.1 or deepstream-9.0. The qt6.12.0-edgeai-plugin-deepstream package depends on one of them, and apt cannot satisfy that dependency when DeepStream is installed by other means.

Setting up the repository

On the device, set up the repository as described in Enterprise Qt Debian Packages. For Jetson Linux, use the arm64 architecture and the noble suite, which is Ubuntu 24. The line in /etc/apt/sources.list.d/tqtc.list is:

deb [arch=arm64 signed-by=/etc/apt/keyrings/tqtc/qt-company-debian-repo.gpg] https://debian-packages.qt.io/debian/enterprise/qtedgeai-6.12.0-arm64 tqtc-noble main

This repository also has the Qt packages that Qt Edge AI depends on.

Installing the packages

Install the Qt Edge AI development package and the build tools:

sudo apt install qt6.12.0-edgeai-dev qt6.12.0-edgeai-plugin-deepstream cmake ninja-build g++

The GStreamer backend of Qt Edge AI uses elements from GStreamer core and from the GStreamer base plugins, such as appsrc, appsink, and videoconvert. The plugin packages depend on gstreamer1.0-plugins-base, so apt installs the base plugins.

The packages install the Qt files in /opt/qt-6.12.0/aarch64-linux-gnu/.

Package contents

The plugin packages contain only the plugin library, in /opt/qt-6.12.0/aarch64-linux-gnu/plugins/edgeai/. They have no headers or CMake files, so you need them to run applications, not to build them. apt installs qt6.12.0-edgeai-plugin-gstreamer by default as a recommended package.

Building the face detection example on the device

The Simple Face Detection example uses Qt Edge AI to detect faces in the video from a camera. Build it on the device:

1. Get the example sources from the Qt Edge AI repository:

   git clone https://github.com/TheQtCompanyRnD/qt-edge-ai.git

2. Configure the example with CMake. Set CMAKE_PREFIX_PATH to the Qt installation:

   cd qt-edge-ai/examples/vision/simplefacedetection
   cmake -S . -B build -G Ninja \
       -DCMAKE_PREFIX_PATH=/opt/qt-6.12.0/aarch64-linux-gnu

3. Build the example:

   cmake --build build

Using Qt Creator on the device

To build in Qt Creator on the device:

1. Install the arm64 version of Qt Creator with the Qt Online Installer.
2. Add the Qt version from the packages. Select Preferences > Kits > Qt Versions > Add, and select /opt/qt-6.12.0/aarch64-linux-gnu/bin/qmake. For more information, see Add Qt versions.
3. Add a kit that uses this Qt version. For more information, see Add kits.
4. Open the CMakeLists.txt file of the example, and select the kit.

Testing the setup

To test the setup, run the example on the device with the DeepStream backend, from the build directory. CMake copies the configs directory to the directory of the executable, and the example reads its configuration file from there, relative to that directory.

cd ./build
./simplefacedetection --provider deepstream --config configs/deepstream/dsdetect1_config.yml

The example uses the default camera of the device. It sets QT_MEDIA_BACKEND to gstreamer, so Qt Multimedia uses the same media framework as the GStreamer backend of Qt Edge AI.

The example shows the camera video in a window and logs the detection results to the console.

Building your own application

Build your own application on the device in the same way. Find the Qt Edge AI module with find_package and link to it:

find_package(Qt6 REQUIRED COMPONENTS Core Qml Quick Multimedia EdgeAi)
target_link_libraries(myapp PRIVATE Qt6::EdgeAi)

Next steps

• Installing Qt Edge AI: other ways to install Qt Edge AI
• Qt Edge AI overview
• Simple Face Detection