Reconstruction system with own images

[TodorBoyadjiev]

Hello everyone,

I want to make the reconstruction system work with my own RGB images and the depth images generated from the RGB images with the model GLPN or DepthAnything.
With the images generated from both models make_fragments.py generates a mesh with 0 vertices.
However everything should go well.

I also have a program that saves frames from a video, generates RGBD images and creates a mesh with open3d from all the frames fragments. The problem there is that there are too many polys because the combined point clouds aren't "cleared" and overlap each other.

I need to get at least one of those solutions running.
Do you have an idea of what's to be corrected?

Thanks!