Open chenritiana opened 2 years ago
Hi. It's intended to be C# class library not a runnable application. I.e. it provides code which you can use in own C# applications. I considered providing a console application with it, but I doubt whether this would be of much use. The resulting voxel grids as mere ply-visualization are not really useful I think. And converting the grids to actual BIM-models of any kind is still work-in-progress.
Hi. It's intended to be C# class library not a runnable application. I.e. it provides code which you can use in own C# applications. I considered providing a console application with it, but I doubt whether this would be of much use. The resulting voxel grids as mere ply-visualization are not really useful I think. And converting the grids to actual BIM-models of any kind is still work-in-progress.
oh,I think It's maybe useful.I'm researching how to use these reconstruction in DT/MR,but the mesh's quality generated by Hololens is quite low,I've seen your mesh in article,the mesh's quality is so good,the grid distribution is regular,so I want to learn the method.One more thing is that how to generate voxel model?Use some algorithm or some software?This probles has been bothering me for a long time.Thanks!
The meshes do not result from VoxIR but are input to it. Actually, the depicted meshes are from the HoloLens (v1). You just need to set the level-of-detail-setting for the mesh observer to fine in order to record meshes of this quality (either in a Unity app with MRTK or directly in a plain UWP C# or C++ app).
The voxel models are created by generating a voxel grid of the patametrized resolution and then just setting all voxels which intersect a triangle of the mesh to not-empty. The not-empty voxels are further classified according to the normal directions of the intersecting triangles. Everything else is done by a cascade of rule-based procedures in voxel space.
The meshes do not result from VoxIR but are input to it. Actually, the depicted meshes are from the HoloLens (v1). You just need to set the level-of-detail-setting for the mesh observer to fine in order to record meshes of this quality (either in a Unity app with MRTK or directly in a plain UWP C# or C++ app).
The voxel models are created by generating a voxel grid of the patametrized resolution and then just setting all voxels which intersect a triangle of the mesh to not-empty. The not-empty voxels are further classified according to the normal directions of the intersecting triangles. Everything else is done by a cascade of rule-based procedures in voxel space.
The meshes in the Figure is just created by set the level-of-detail-setting for the mesh observer?It looks good.Maybe I shoule modify some key parameters.
Can this repository run independently?Or run with Unity? How can I use it?