Hello everyone, I am new to WebAR and have the following issue:
I am trying to create an AR add-on for a commercial that sets the initial alignment of my scene based on an image target, but then keeps the scene alignment with world effects once the image target disappears (once the ad is over). Unfortunately, I have very minimal previous knowledge and so far couldn’t piece a solution together from the documentation.
So far I couldn’t find any similar approaches online, which lead me here.
Any help or resources I can use to learn are appreciated!
Hey Marcel,
This project seems like it does what you’re looking for: Image Target SLAM Model (A-Frame). It sets the initial alignment with an image target and continues using world effects after the target disappears.
If you’re new to 8th Wall, this example is probably the easiest to start with. While VPS could work depending on your scenario, this approach should help you get started quickly.
1 Like
Hi George! Thank you so much for your response! I‘ll try it first thing tomorrow!
1 Like
Unfortunately, I am still running into many problems with alignment and overall tracking.
When testing the target itself the object position and alignment work very well, but in practice, the object loses its tracking a lot or does not align with the target in the first place.
I’ve considered looking into VPS for specific outdoor use cases, but the experience is intended to work in any living room essentially, which rules that out too.
Is there any example project that aligns the object similar to the bounding box used in target testing that also uses SLAM I could refer to?
I’d take a look at our video here that outlines some best practices when working with image targets. Feel free to share your image target, and I can take a closer look to highlight any potential issues that might be affecting performance.
Thank you! I’ve managed to work out a majority of my issues regarding the image tracker. Turns out that all it took to fix the major flaws was for me to realize detail.rotation was stored as quaternion instead of XYZ degrees like in HTML and detail.scale being a single value for image targets, which took me waaaay too long to test
I’m hoping that most of the other issues with tracking can be resolved by working out absolute scale with the coaching overlay. Regardless, thank you for your help and quick responses!
1 Like
As I’ve also commented on this post I am now running into major issues with world tracking, which seems to lose track very easily…
Even in an ideal scenario with my tracker on the wall, printed on paper in a size that works great with regular image tracking, once I introduce SLAM tracking, the issues keep stacking up.
Since the target for the project will inevitably have to deal with some suboptimal conditions (screen glare, TV stands, or tables throwing off the track) I feel like I might not be able to finish the project…
Is there anything else I can try or is this simply an engine limitation?
Edit: The issues I described aren’t exclusive to either absolute or relative scale.
Hi Marcel,
Would you be able to share a video of the issue, along with the URL of the experience? If needed, I’d be happy to arrange a meeting to help triage the issues.
Hi George! I’ve recorded some examples, unfortunately, I could only get references in our office environment, I’ve uploaded them to YouTube unlisted and linked them here since the file was too large for the forums. I’d be happy to arrange a meeting if possible, my timezone is CEST (UTC+2), perhaps we could do some time next week?
Edit 1: forgot to add the link to the experience
Edit 2: from what I gather this post outlines a very similar if not identical issue to what I am experiencing on my end, documented very well. It seems like the issue has yet to be resolved there as well.
Thanks in advance and have a great day!