Can this repo support mesh extraction given multiple view depth-maps?

[hbb1]

Great works. I wonder how can this method extend to multiple view depthmap? Can I use it for unbounded mesh extraction in a fashion like TSDF fusion?

[mazeyu]

I think it is possible but the method to reduce the noise in the depthmap is out of the scope of this project. Assuming the depthmaps are perfect, I could work on providing such interface later.

[hbb1]

Thanks for your reply! Assuming I have some perfect RGBD images, I would like to obtain a mesh for unbounded scene. Traditional volumetric fusion-based methods allows fusing them into a global TSDF but scales badly for large scales when meshing, while your proposed method seems to be an awesome solution. If I want to extend your method myself, what files should I look in particular? Thanks.

[mazeyu]

I recommend constructing your own SDF function based on depth maps and wrapping it as a function to feed to the "kernels" argument in ocmesher/core.py. (kernels is a list of functions, in your case it can just be a singleton)