MR QRcode Detection.

this project for example (very high budget project) had to rely on a separate outside in tracking system to reorient and correct drift in their mostly fully VR roomscale project (which then had a passthrough component near the end)

https://phi.ca/en/studio/works/blur/

Here in this picture you can see the tracker Leds on the floor and on each headset themselves hooked up to an optitrack (that their PR person didn't photograph)

This low(er) budget dance piece I saw in Prague has a real dancer moving and dancing with their volumetric video "replica" in passthrough

https://www.displace.org.uk/replica/

 and this one required a special onboarding headache (recenter while facing the correct way and standing on an X and quickly put the headset on a viewer before it resets itself) and thats so the performer and their volumetric video representation (which they cant themselves see) is hopefully aligned for the audiences.

Projects like these would greatly benefit from using something like a QR code or other real world scene reference for alignment, but the current api seems really geared towards just reading a code on a phone held up arms length from the passthrough camera instead of a simpler tracking reference poster sized printed on the floor or wall across the room.

This feature isn’t really designed for real-time tracking or spatial alignment. It’s mainly meant for reading QR codes as data carriers for example, login links or other static information, not for anchoring or tracking in mixed reality spaces.

In general, I wouldn’t recommend relying on any kind of QR code for tracking or alignment purposes. They’re not optimized for that use case and tend to perform poorly at distance or under varying lighting conditions.

If your goal is to establish stable visual references in space, ArUco markers would be a much better choice. They’re designed specifically for computer vision tracking, much easier to detect and estimate pose from, and more robust for scenarios like real-world alignment and drift correction in MR setups.

If you decide to go in that direction, I definitely recommend scanning them using OpenCV  it’s reliable, flexible, and works well across different environments.

If this answers your question, please mark it as solved.

~ Nigel (XR Bootcamp)