Performance issue of the extended tracking in image marker

Hi there!

I’ve question about the performance issue of the extended tracking in image marker. I want to create the effect that after scanning the image marker, the model will anchor at the world position. And even the marker is lost, the model will still appear at the correct position.

In SparkAR, the fixed target tracker somehow can do the effect, but it is not very stable that you must slowly move away from the marker.

On the other hand, the plane tracker have the world anchor function which can lock the model in the world space. So I thought to combine both trackers, which to get the image marker position and anchor the model using plane tracker. However, the target tracker and plane tracker can’t be in the same project.

Therefore, is there any method to increase the performance of the fixed target tracker or other solution to do the image marker extended tracking? Thank you!!

Ps. This is a work of Takashi Murakami at The Broad Museum which smoothly perform the extended tracking in image marker. And it should be created by the Buck Design. Any thought of how they can make this effect?

https://www.instagram.com/ar/1373277393083432 (the filter seems can’t open now)
https://www.instagram.com/reel/CixuzjbJLpI/

Perhaps they just fix the position of the objects after finding the marker. (global transform and other patches)

Spark cannot use the trackers you named at the same time

Yeah, and currently find out that SparkAR is initializing SLAM after finding the marker. However, after replace a new marker in the demo project (3D Animated Poster), the SLAM initialize failed and the extended tracking failed too. So any method to control the SLAM function?

mmh, checking the video it might be that Meta gave them access to a better version of the tracker that’s still in BETA. Seems like a big enough project for them to do it