Cytoskeleton

Comment

Comment by Jake Hebbert on January 29, 2014 at 3:52pm

I played around with the idea a little this week and think it is over my head right now.  Atleast without dedicating more time than I have available right now.  Thanks for the suggestions of directions to go in though! :)

Comment by Daniel Piker on January 22, 2014 at 9:57am

Jake - Yes I see now, and think it would be possible to extend a similar approach to the kind of line networks you show.

However you do need the lines at each node ordered radially, as Dave points out, and also a normal vector.

Once you've got one of these, you can use it to get the other.

I would suggest taking a point a short distance away from each node along each connected edge, and getting a best fit plane through these points. You can then sort the connected edges radially in this plane, and use its normal.

That still leaves step 6 though - I guess you'd need to interpolate the normals between the ends of each curve. I think SpiderWeb could come in useful in all of this.

Comment by David Stasiuk on January 22, 2014 at 9:35am

Three curves connecting at any given node should be pretty straightforward...once you start to get more than that you need to sort them correctly, which becomes problematic. That's where the hulling operation in exoskeleton gets used...but of course the connections it creates are irregular.

Comment by Jake Hebbert on January 22, 2014 at 9:05am

I should've explained more clearly:)

For awhile now I have wanted to make a component that allows me to pass in a bunch of curves some of which are connected. What I want out is a good looking mesh like the one you get from your cytoskeleton. The trouble I have had in thinking about it is I can't figure out how to make the joints (I want cleaner more simpler joint than the exoskel gives.)

So your picture gave me a little insight into how it might be done...

Comment by Daniel Piker on January 21, 2014 at 3:10pm

Jake - not sure I follow.

You mean to apply it to general curve networks?

This approach relies on the input having a mesh topology (though as it is Plankton, this can be an ngon mesh).

(Exoskeleton doesn't have any such requirement, and can be applied to any line network, but it outputs triangulated meshes instead of clean quads).

Comment by Jake Hebbert on January 21, 2014 at 12:48pm

I was thinking about this and would there be a way to use this method to create a curve skeleton?

1. pass in a set of curves.

2. find where the curves meet

3.  create a mesh node at each intersection 

4.  divide the curves to get polylines.

5.  connect the mesh nodes via the polylines.

Does that make sense?  

Comment by Manuel Sotomayor Millan on January 21, 2014 at 12:05pm

wow

Comment by Mihai Pruna on January 21, 2014 at 11:56am

Awesome job!

Comment by Daniel Piker on January 20, 2014 at 1:36am

Mateusz - good point. I might have to adjust the way I'm calculating the normals, but yes, I think if the underlying mesh is CP, circular, or conical, then these quads can be planar.

Comment by Mateusz Zwierzycki on January 19, 2014 at 6:25pm

aren't those (bottom image) quad faces planar ? (as a "side effect" of producing CP mesh ?)

• ‹ Previous
• 1
• 2
• Next ›
• Page