we have some problems with AR Recognition.
We have made different tests with our markers and also with some markers we had taken from your sample Unity project. (Wikitude 8.6.0) [WikitudeUnityExample]
For this example, we had selected a 3 stars Diplodocus.
The first problem is the scale of the 3d objects. The air conditioner size is 798x295x189, while the marker size is 91,7 × 126,7 mm, but when when the marker is recognized the 3d object appears little than the marker (The parameter "target height" in wtc editor also seems not doing anything). So we had to manually scale it of an arbitrary factor dependent of the marker size.
The second problem is that the 3d object is not stable, sometimes is not aligned with the marker and flickers expecially when extended tracking is enabled.
We provide the wetransfer with the movies. There are three movies inside,
· the first is recognizing the marker on a a4 format without the extended tracking
· the second is recognizing the marker on a a6 format without the extended tracking (our client would prefeer to use this instead of the a4)
· the last is a4 with extended tracking enabled (which happens to seems bugged sometimes and did not helped in tracking the airconditioner in a stable way)
This was experienced with multiple phones (S7, S6, Google Pixel XL) and multiple unity versions (2017 , 2019)
Regarding the first problem, we're aware that the way objects are scaled is not intuitive at the moment and we're working on changing that.
For now, an image target will always be 1 unit in height in Unity, regardless of the target height parameter in the wtc file. The target height parameter is currently only used when calling APIs to compute the distance to the camera or the distance between two image targets.
The preview button from the ImageTrackable will show an accurate depiction of the target in the Unity scene, which should help you align the augmentations more easily.
Hopefully, we'll be able to roll out a new version that works as you're expecting soon.
As for the second problem, would it be possible for you to send us the project, as shown in the video, so that we can try it on our own and see how to best improve the tracking quality? You can send it at our support email, if you'd rather not share it in the forums.
One thing I will note is that extended tracking won't help very much in the scenario you have in the video because there are no feature points around the target. It could also explain why it's behaving worse than regular image tracking.