[BUGS]: return_collapses

[zhuyuanxiang]

Environmental: python=3.8.19, vedo=2024.5.1, vtk=9.3.1, fast-simplication=0.1.7 os=Windows11 mesh's vertices number > 50000, cells number > 100000 Code:

        mesh = vedo.file_io.load(file_name_path.as_posix())
        vertices, cells = mesh.vertices, mesh.cells
        vertices_num, cells_num = len(vertices), len(cells)
        vertices_out, cells_out, collapse_index = fast_simplification.simplify(
            vertices, cells, target_count=20000, agg=3, return_collapses=True
        )
        vertices_out_num = vertices_out.shape[0]
        collapse_len = collapse_index.shape[0]

Then, it is a error that collapse_len <(vertices_num-vertices_out_num), because collapse_len's meaning is how many points is collapsed, and collapse_len should greater than (vertices_num-vertices_out_num).

Now, I need use this index for collapse the labels, please give me some help! Thanks!

[akaszynski]

Pinging @Louis-Pujol

[Louis-Pujol]

Hi @zhuyuanxiang,

During the decimation, there is indeed two operations :

• collapses, that are tracked by collapse_index
• cleanings as certain points can become isolated after some collapses and must be removed from the final mesh

If you look at the replay function implementation, after the collapses are redone, we clean the mesh by removing isolated points.

Then, collapse_len is not exactly the difference between number of points before and after decimation but rather a minorant of that number, so having collapse_len <(vertices_num-vertices_out_num) does not sound wrong to me.

I have written higher-level interface around fast-simplification for another project (scikit-shapes), the possibility are illustrated in the Multiscaling section of the gallery of examples.

You can have a look at the code for the Decimation and Multiscaling classes to figure out how fast-simplification is used there.

[zhuyuanxiang]

Hi, @Louis-Pujol

Thks for your help!

Now, I know the reason about collapse_len <(vertices_num-vertices_out_num), but I need the isolated_points in the _map_isolated_points() for collapse the labels in the pointdata. I hope how to do it.

If I could get a new simplify() to get the isolated_points, or I could use the replay_simplification() to simpily the mesh and modify it for return the isolated_points?

Maybe I know how to do it! I found the test_relay.py in the project and get the method to obtain indice_mapping which can map the all vertices to the decimated mesh, so I will try to use this function and reply the result ASAP.

BTW: I have not found isolated_points in another project scikit-shapes too.

[Louis-Pujol]

Hi @zhuyuanxiang,

replay_simplification() has been designed to perform both operations: concatenation and cleaning of isolated points. The output consists of three arrays:

• The decimated points
• The decimated triangles
• indice_mapping with the following description: indice_mapping[i] = j means that vertex i of the original mesh was mapped to vertex j of the decimated mesh.

So, if you use replay_simplification(), you don't need to worry about isolated points, as they are just an intermediate step in the global decimation process. I think indice_mapping is the output you will need to collapse the labels. In scikit-shapes, we use it to transfer data (landmarks or signals) from fine meshes to coarse meshes.