Unrolling corset pattern segments with Squish

Im getting accurate edge lengths using the custom preset bndstr 10. bndcompress 10. intstretch1. Intcompress 1
But am getting many jagged edges. How do I fix this. Model below.
NEW Corset mannequin.3dm (7.0 MB)

Im my rush I forgot to add the magic words Please and Thank you.

Here are my results with all deformation values set to 10. Jagged edges eliminated, but im still out on the longest curve by 2,7 mm, but Im sure I can fix this by altering the settings. Although ti would help a lot if they were explained properly somewhere and I wasnt just geusing.

Im trying to get all edge lengths under 1mm difference.

Hi Simon - you might try squishing a mesh version of the panels - the toll uses a mesh anyway to squish, and if you mesh the thing yourself ahead of time you can control the mesh better - try something like 256 minimum initial quads (detailed controls in the Mesh command) and an aspect ratio of 1. Does that make anything better?


Thanks for the input. I will try that out later.

Na thats a little furthur out. Id like to skip the mesh step and focus on the deformation values if at all possible, as my client want to build patterns from the form I have made, so I want to avoid complicating steps if at all possible.

I have all values set to 10 at the moment, which is soo very close to being perfect, only 1/2 a cm expansion on a 41 1/2 cm edge.

So logicaly im assuming I need to reduce the stretch values only (when i made all deformation values 5 it made the longer curves more expanded, so i dont quite understand how these work)

Yeah int/bnd stretch set to 5 and int/bnd compress set to 10 doesnt help either.

The thing is, the finer and more accurate the mesh, the more likely you’ll get a polyline on the unrolled shape that matches your input edge. Here, squishing a nice fine mesh makes the result slightly but noticeably closer:

Input edge: 41.097
Squish edge: 40.932
Squish finer mesh version: 40.995

You can get an idea of the accuracy difference on these shorter curves:


Ok thats interesting. The shorter edges are not the problem realy, its the longer more curved ones. I will experiment more with the mesh settings then.
Thank you.

Ths seems to be the most dense I can make the mesh.
And its still furthur our on the longest edge than when I squish straight from the surface. I assume im using the mesh settings wrong.

PS: Arn’t you collecting squish data like this? (im sure I read that somewhere) Id be more than happy to send all our results to you.

Hi Simon - set minimum quads higher and max edge length to increase the number of polys. I’m not saying that is the only factor, but I think you want to give the thing a fairly accurate representation of the edges as polylines. I’ll let the developer know about this thread - I think the settings are partly in the realm of the magical, and not easily understood by mortals other than by experimentation - though there is a fair amount in the Help file about what the settings do.


1 Like

Thanks dude. I realy appreciate the advice.

I’ve been watching this thread and it’s an interesting problem. It’s been bouncing around in my noggin and I’m thinking of a different work flow that a) might give very accurate dimensions, and b) might be more trouble than it’s worth.
It might be re-inventing the wheel, but here goes. I’ll consider one segment that you’ve already imposed upon the mannequin, say the right panel, the one that has the right edge that would go through the belly button.
Duplicate Edges, so you have 4 separate edge curves
Project to C plane
Run Tween Curves between opposite pairs. Number of curves to be representative for each pair, e.g. divide the shorter edges into, say 12 curves, and divide the long pairs into 80 curves (I’m just making up the numbers, so bear with me).
You’ve essentially made a grid, of sorts.
Project the grid on to the mannequin’s surface. If some of the curves don’t extend far enough to go to the edge of the pattern on the model, undo, extend the curves that are too short and repeat the Project operation, or maybe Extend Curve on Surface?.
Trim the projected curves with the outline (border) curves, if needed.
Measure the length of the curves that have been projected on to the mannequin’s surface, and that should be the exact dimension for each curve in the grid.
Adjust the length of the planar grid curves, both horizontal and vertical, to match the measured lengths of the projected curves.
Put Points at the ends of each curve.
Curve through Points.
Surface from Edge Curves.

At this point, shouldn’t the flat pattern be dimensionally accurate?

Since the right edge of the pattern (the one that would bisect the belly button) would be reasonably straight, I’d probably adjust those line lengths to the left. For the other direction, I’d probably take some test measurements of parts of the long lines to see where the greatest dimensional changes are localized and make a decision about adjusting line lengths based on that. Probably more deviation in line length going over the bust, so picking a line along the greatest height and making a proportionate adjustment in length from there.
I imagine the repetitive grunt work can be scripted out. And, because I don’t know how Squish really works, as I said at the top, this might be reinventing the wheel, but I’ve been thinking that the process I described above should give really accurate results. What do you think?


Slurps coffee.
Reads with interest.
(I will get back to you…thanks)

Thank you for your input Doug, but that is far to complex to be efficient with making patterns.
Im pretty sure ive cracked the mesh settings now.

Its just the squish settings that need refinement (im getting the best results with all deformation values set to 10, it would REALY help if I knew what squish set them to by default so I had a point of referance)

I know I am probably asking a lot of it, and simplifing the mannequin surface curve network before splitting into panels may make the unroll more accurate.

Hi Simon,
I don’t know much about Squish but used the meshing tools quite a lot…
Your settings works well with this surfaces but could yield to unexpected results as they are.
Try something like that:

Entering zeros tell the mesher to not pay attention to this setting.
This way you can tell almost exactly how many polygons to create.

Good tip Marc.
Thanks :smile:

HI Simon,

Thank you for taking the time to work through this example.

Case like this corset where the original surface is not developable ( https://en.wikipedia.org/wiki/Developable_surface ) are generally not handled well by squish.

In your corset case, It appears that once you get an appropriate mesh, you still need the ability to specify that some curves on the model should not have their lengths modified and are willing to have other areas subject to more distortion. Is my understanding correct? If so, I will file a bug for the Squish command with your model as an example.

– Dale Lear