Johnathon Selstad

The dumbest possible Minkowski algorithm I know (for a convex “tool” object) is to go through each face of one mesh, hull the other mesh to each of the vertices...

I think there's a simpler (if slower) solution to offsetting that preserves manifoldness and adds fillets... ```py def offset(a, distance, resolution=20): a_mesh = a.to_mesh() sphere = manifold3d.Manifold.sphere(abs(distance), resolution) # Hull...

It might be nice to do Linear Sweeps too. I wrote a small sweeping implementation using [CoACD](https://github.com/SarahWeiii/CoACD) to see how bad it would be:  ```py # Linear Sweep import...

@pca006132 > After this we still have another problem: How should we compute the "blend surface" by subdivision? Again, for the convex case it is simple, just add the two...

One quality boost that just occurred to me is to rotate the hull spheres so that one of their vertices align to the mesh’s surface normal. This should be a...

I’ll admit, the only non-convex/non-convex sum I’ve ever wanted to do is a toroid sweep along a spline path (and there are far more efficient ways of doing that). My...

I think approximate methods are slower because they are iterative, and aiming for the smallest number of hulls possible. I’m hopeful there are fast and exact methods (like this guys’s...

I have a new dumb algorithm in mind for convex decomposition: - Find the most concave edge in the manifold - Split the manifold by a plane along this edge....

Bisecting plane sounds good, but how do you calculate it for the swept surface? > Interesting. These ideas remind me a bit of binary space partitioning trees (BSP), which are...

Hrmmm, testing out my Solid Exact Convex Decomposition, it seems like it takes about as long as CoACD and VHACD, but outputs 2-6x more convex parts (close to the order...