args.target_edge_len ignored when the edge length is large

[ikoruk]

There seems to be a threshold over which the mesh is simply returned without any refinement, ignoring target_edge_len. This seems to be the case if the edge length is even 0.01 the bounding box size.

[Yixin-Hu]

Hi, the target_edge_len controls the max edge length in the output. This is because if the input has high-frequency features and you want to preserve them, then it's impossible to get a very coarse mesh.

[ikoruk]

Hi, the problem is actually the opposite. Above a certain edge length the mesh returned is always the same mesh. (Which is too coarse)

[Yixin-Hu]

Could you share the input and the command you used?

[ikoruk]

I'm trying a very simple triangle: [0, 0], [1, 0], [0.5, 1] with a target edge length of 0.05. It results in an average edge length of 0.2847.

Here is the output:

Loading and preprocessing ...
remove duplicates:
#v 3->3
#e 3->3
simplify:
#v 3->2
#e 3->1
Loaded and preprocessed.
time = 0.0598474s

BSP subdivision...
#v 102->125
#f 166->212
BSP subdivision done.
time = 0.00156363s

Mesh optimization...
#v = 125(125)    unrounded = 0
#t = 212(212)
max_energy = 10.2876
avg_energy = 2.89565
min_angle = 12.7835, max_angle = 116.565
< 5: 0, < 10: 0, < 15: 0.0188679
> 175: 0, > 170: 0, > 165: 0

//////////////// preprocessing ////////////////
edge collapsing...
queue_s.size() = 595
success 30
re-push 68
queue_s.size() = 65
success 4
re-push 6
queue_s.size() = 1
success 0
re-push 0
edge collapsing done!
time = 0.475119s
#v = 91(125)    unrounded = 0
#t = 148(212)
max_energy = 3.69612
avg_energy = 2.33998
min_angle = 24.4348, max_angle = 109.722
< 5: 0, < 10: 0, < 15: 0
> 175: 0, > 170: 0, > 165: 0

//////////////// postprocessing ////////////////
edge collapsing...
edge collapsing done!
time = 6.2761e-05s
#v = 91(125)    unrounded = 0
#t = 148(212)
max_energy = 3.69612
avg_energy = 2.33998
min_angle = 24.4348, max_angle = 109.722
< 5: 0, < 10: 0, < 15: 0
> 175: 0, > 170: 0, > 165: 0

Mesh optimization done.
time = 0.476664s

A target edge length of 0.02 results in an even higher average edge length of 0.33. Only at 0.01 does it return a closer (though still far) average edge length of 0.0349.

[Yixin-Hu]

Ok, I'll check it