[Alpha version] Feedback needed for feature points based meshing solution (AR Foundation)

Hi,

In the last few months I was working on the feature points based meshing solution for ARCore and ARKit and I'm ready to release alpha version. With spatial mapping app developers can create a mesh and enable occlusions. Missing that feature motivated me to create it on my own.

Right now there are still improvements needed, especially around merging meshes and perf and I hope I will finish it in the following weeks.

Here how it looks like (ARKit, ARKit with screen recording and ARCore):

It's all Unity project build on top of the AR Foundation. Runs on Android and iOS. Solution doesn't use any external libraries.

What would help me prioritize the work, is to get a list of important things from you. I would really appreciate if you could share with me some answers to what is important to you:
- Quality of the mesh or bigger mesh?
- Are you fine with some holes or you need mesh to be perfect?
- Do you care about mesh with texturing, vertex color or just occlusion?
- Is it important for you to be cross platform?

To help with the distribution:
- Are you individual or a company?
- Are you building games or apps (retail or enterprise)?

You can reply here, priv, use email contact@ARMeshing.com or Twitter. All is fine.

Official website: http://ARMeshing.com/
Twitter: https://twitter.com/BinaryBanana

Thanks!

 

@BinaryBanana Looks pretty good. I would be mostly interested in mesh generation around the area where a touch event occurs so you can move objects around by raycasting against the generated mesh. It would also run a lot faster than trying to generate a mesh of an entire space.

Ar you aware that https://www.6d.ai/ is also working on occlusion and mesh generation? TorchAR is already collaborating with them https://www.torch.app/.

https://www.torch.app/blog/expanding-augmented-reality-creator-platform-6d-ai-torch

 

Getting a mesh around touch events might be tricky as you might point at something at a greater distance for which you won't get feature points from the place where you standing or points accuracy will be low. Thus I believe you need to build complete map and update it over time.

Yea, I know 6d.ai. They are building really nice solution with great results. The main difference though is that they are not working on top of the ARCore or ARKit, they are a solution on their own.

I personally feel that not every app developer will feel comfortable to base the entire project on 6d.ai or aboundlabs (the other, similar one) and some will prefer to stay on ARCore or ARKit, where I hope I can provide alternative. At least until Google or Apple will include it in their platform.

 

Hey any news on this? It looks very promising.

 

Thank you. Unfortunately we are not working actively on the project anymore. We believe it should be and soon will be part of the platform (ARKit and ARCore) and for that reason we put the project on hold, at least for now.

 

I understand, there is any chance we can get the SDK so we can test around until that day comes when ARKit and ARCore will implement it?

Cheers

 

There's a module in AR Foundation called ARMeshManager. Does that have anything to do with this?

 

Well depending on which platform you use it will work or otherwise return null.

If you are trying to use it with ARKit or ARCore, like I tried it will not work. Answer here.