Reset of pivot point.

[gdemme]

It should be possible to change the pivot point of the geometry.

[wherget]

You can change the pivot point of the 3D view to "pivot around the whole scene" and "pivot around a specific object". (Meaning the center point of the respective bounding box.)

We chose these two in our first approch to a pivot camera as we figured these would allow you to inspect from a decent number of viewpoints.

Can you describe in closer detail how you would want to move around the scene, and how the camera should behave? Maybe your Scenario setup is too coarse-grained and you want to have COMPASS handle finer grained structures of your geometries' hierarchy, such that the previously-mentioned single object pivot provides more use to you?

[gdemme]

I am referring to the pivot point of objects.

If the pivot point of the object lies outside the object, scaling and rotating is weird. It would be great if it is possible to move the pivot point of the object relative to the object's geometry.

One work around is to put the object into an empty scene node and use that scene node as pivot point. But it's a little cumbersome.

[wherget]

Do you have sample geometry where this occurs? I'm having a hard time thinking of a solid object where this would be a problem. A horseshoe, or half-donut maybe?

Come to think of it, using your workaround the other way round, i.e. adding an empty node below the geometry as a manipulable pivot point should work as well and be more stable w.r.t. moving the geometry around the hierarchy. In a setup like that the pivot node's transform is relative w.r.t. the geometry it belongs to.

How would you expect to manipulate the pivot point? Ideas we've thought of/seen are: a nub that snaps to the surface of a geometry, indicating the new center point; you could also "move" the pivot point on a plane perpendicular to the viewing axis, e.g. using a modifier key. The first is probably more reliable, but harder to implement. Do you have other ideas?

[wherget]

A quick follow-up to this: When you initially switch into inspect mode, the pivot point stays at the look-at point of the free-flight camera (our rationale at the time was not to make the camera jump when you switch modes). This may result in the behaviour you're seeing, i.e. the pivot point being outside the geometry. Once you click the "Reset Camera Position" button (up-arrow symbol), the pivot point should be correcly centered on the object. Can you please check whether this behaviour was causing your issue? If so, would you like to suggest a more intuitive workflow, since our choice apparently caused confusion?

[gdemme]

Sorry for the confusion.

Each object has a pivot point, where the transform widget for that object appears.

If that point is too far off the actual geometry, transformations get difficult. E.g. if you center to the object in inspect mode, the widget will be outside of the view.

I think, that this problem is more due to bad object modelling and not actually a problem of compass. But having a possibility to cope with bad models would be great.

[wherget]

Oh, I see what you mean now. We usually refer to that point as the "origin" of that geometry, hence my confusion. I agree that this is mostly a modelling problem. We usually take a "Garbage in, Garbage out" stance towards problems as this, so I expect your proposed feature to be rather low priority in COMPASS. However, it very much sounds like a feature that would fit very well for ATLAS.