The Qt Group has been working on various AI integrations for the Qt toolkit in recent times like a QML profiler skill for agentic development to help with performance profiling. They have also worked on Qt AI coding assistants and AI agents in the Qt Creator integrated development environment. Their latest AI use is now preparing to introduce a new module in the Qt toolkit itself for edge AI inferencing.

Qt developers are preparing to introduce Qt Edge AI as a new sub-module for developers to create edge AI applications. With the Qt toolkit usage this would be done in an AI runtime agnostic manner with initial support targeting embedded devices like NVIDIA Jetson/Thor, Qualcomm IQ, and NXP i.MX95 platforms.

The initial Qt EdgeAI implementation is described in the Qt code review ticket as:

"Add first version of EdgeAI focusing on vision AI

The commit adds a module with C++ APIs for creating a pipeline of inference steps and running machine vision tasks, such as object detection and classification. An interface is defined for providing platform/runtime tied implementations for the step.

The API depends on QtMultimedia to receive QVideoFrames, and a private QtMultimedia API to further access the underlying GstBuffer so it can be fed to the inference pipeline."

The initial Qt EdgeAI sub-module was co-authored via the Cursor AI agent.

Follow-up code review tickets add an NXP NNStreamer back-end plug-in and an example application of using Qt EdgeAI for face detection. We'll see in time what more is tacked onto EdgeAI beyond the initial vision AI focus.

This mailing list thread from the Qt developers lays out their plans for adding this EdgeAI module to the Qt toolkit's arsenal.