Example: for CircularLoopWholeSpace, you get the same responses whether orientation is [0,0,1] or [0,0,-1] (when they should be mirrored).
Interestingly, if you use just a slightly offset orientation (ex: [0,0.01,-np.sqrt(1-.01*.01)])), you get the right answer for the pointing down orientation case.
I narrowed it down to this line of code, an issue with the in-house spatial.rotate_points_from_normals
I think the best course of action would be to replace the existing custom, in-house tooling, with scipy rotations
Example: for
CircularLoopWholeSpace
, you get the same responses whether orientation is[0,0,1]
or[0,0,-1]
(when they should be mirrored). Interestingly, if you use just a slightly offset orientation (ex: [0,0.01,-np.sqrt(1-.01*.01)])
), you get the right answer for the pointing down orientation case.I narrowed it down to this line of code, an issue with the in-house
spatial.rotate_points_from_normals
I think the best course of action would be to replace the existing custom, in-house tooling, with scipy rotations