What's the point of defeaturing?

Maybe I misunderstood something, but I thought that defeaturing was an optimization?

But it keeps pretty much 90% of the polygons:

Now, since this is done AFTER tesselation, I get why it has to be like this, but if it was done BEFORE tesselation, it wouldn’t have to.

99% if not 100% of all holes made in CAD software are made with trimming, and that retains the surface data where the hole is placed. You can always recover the full surface after you’ve made a hole, and most CAD software has an “untrim” feature.

If you want to see an example of a good defeaturing implementation, look at Ansys SpaceClaim.

Can I bump this?

Does anyone here currently use defeaturing in Datasmith, and if so, why?

Thanks for the feedback. Agree that working on NURBS data can produce better results in many cases.
We have many users that get data already tessellated and need to remove mechanical form features on triangle meshes, that’s why we have the defeaturing technique working on meshes.
In order to have better gains, it’s often recommended to use decimation afterwards. In the example above, the mesh reduction tool should remove most of the coplanar triangles without visual quality loss.

Thank you for the reply! I wasn’t aware of the “mesh reduction tool” you mention. Usually, for CAD imported surfaces, this destroys the custom split normals that tesselation produce, so if Unreal’s tool doesn’t do this, that’s an improvement.

Since I work with double-curvature surfaces which can still have holes in them, I’ve switched to defeaturing in Rhino for now.