+- tyFlow Forum (https://forum.tyflow.com)
+-- Forum: tyFlow Discussion (https://forum.tyflow.com/forum-1.html)
+--- Forum: Feature Requests (https://forum.tyflow.com/forum-4.html)
+--- Thread: Can you add the "matrix" menu to the "rotation_align to shape mesh (oriented boundari (/thread-3229.html)
Can you add the "matrix" menu to the "rotation_align to shape mesh (oriented boundari - cgdvs - 08-07-2022
HI,Tyson!
Can you add the "matrix" menu to the "rotation_align to shape mesh (oriented boundaries)" operator, as shown in the figure, after the particles use this mode, the axial direction will appear in both positive and negative directions. If there is a matrix menu, this problem can be avoided. thank you
RE: Can you add the "matrix" menu to the "rotation_align to shape mesh (oriented boundari - tyFlow - 08-07-2022
Well the oriented bounds are determined by constructing a principle component analysis on the shape's vertices, which naturally results in flipping because there's no clear 'correct' direction of a point cloud. In many cases where shapes are oriented in random directions, being able to specify a target direction wouldn't make sense...you happen to have a use case where it would make sense because all of your shapes are flat. So...it wouldn't make sense to add that kind of 'forward vector' specification in the same way, but it could make sense to be able to specify some kind of directional preference, where the TM will be flipped if the dot product between the two is less than 0 (vectors point away from each other). It could also be interesting to add a setting that will bias the PCA matrix towards the world XYZ matrix, flipping any axis that diverges more than its mirror axis....that may even cover more use cases than just specifying a favored up axis. I will consider adding that.
RE: Can you add the "matrix" menu to the "rotation_align to shape mesh (oriented boundari - cgdvs - 08-08-2022
(08-07-2022, 06:37 PM)tyFlow Wrote: Well the oriented bounds are determined by constructing a principle component analysis on the shape's vertices, which naturally results in flipping because there's no clear 'correct' direction of a point cloud. In many cases where shapes are oriented in random directions, being able to specify a target direction wouldn't make sense...you happen to have a use case where it would make sense because all of your shapes are flat. So...it wouldn't make sense to add that kind of 'forward vector' specification in the same way, but it could make sense to be able to specify some kind of directional preference, where the TM will be flipped if the dot product between the two is less than 0 (vectors point away from each other). It could also be interesting to add a setting that will bias the PCA matrix towards the world XYZ matrix, flipping any axis that diverges more than its mirror axis....that may even cover more use cases than just specifying a favored up axis. I will consider adding that.
You're right. You can't simply add an attribute. I think if there is no better way, the use of reference objects can also solve some of the problems.