Is a good way to intuitively think about hierarchical rep is to imagine ur object like a marionette? Such that if you rig all the objects that make up an arm like in the following slides "manipulating one string" will move the whole arm? Is it possible for "strings to get crossed" when traversing the grouped tree? If so what can I do to prevent additional complexity while still retaining the high-level editing? (aka when is it actually needed and less of a possible headache?)
and why shouldn't I just simply group objects together to apply the same transformations to a multiple objects? My instincts feel the need to default to this if I a going to have to traverse a tree anyways and a solid answer on why this isn't a strong development idea for lets say... a game engine, will help cement my understanding on the use of heirarc. rep.
danielhsu021202
A way to think about this is puppets! Say we want the puppet to raise its hand. We just pull the string associated with that arm, and the hand raises. We didn't have to think about each individual joint and body part involved in raising the hand.
Here, the linear representation would require us to consider every part on its own, aka think about every joint and body part, whereas in the hierarchical representation, we just lift the hand and everything else (arm, elbow, forearm) raises itself.
sylviacx
is it appropriate to think of a parent group as defining a coordinate system, or planes, or some kind of space that the subgroups apply themselves to? So when the parent transforms, the subgroups just apply their transforms in the new coordinate system?
adam2451
When applying transformations to leaves in the hierarchy tree, is there a way of limiting transformations such that they make sense in relation to the parent nodes of the tree? Or would we have to be conscious of that when applying transformations ourselves. For instance, it would be awkward if we made the hand on an arm accidentally slide 10 feet away from the arm it was supposed to be attached to.
caelinsutch
In practice how many groups or trees do models have and when calculating the "end-state" of the tree are there optimizations for caching / speeding up calculations of relative coordinates for each step?
sebzhao
So is hierarchical representation relational in the sense that groups are defined with transforms relative to another group (and is that why for example if you raise the hand it would raise the arm--because the arm is defined as a transform relative to the position of the hand?)
Alescontrela
@sebzhao I think this is indeed the case. Each new leaf in the hierarchy is defined by features in it's local frame, as well as a transform from the parent frame to the local frame.
Is a good way to intuitively think about hierarchical rep is to imagine ur object like a marionette? Such that if you rig all the objects that make up an arm like in the following slides "manipulating one string" will move the whole arm? Is it possible for "strings to get crossed" when traversing the grouped tree? If so what can I do to prevent additional complexity while still retaining the high-level editing? (aka when is it actually needed and less of a possible headache?)
and why shouldn't I just simply group objects together to apply the same transformations to a multiple objects? My instincts feel the need to default to this if I a going to have to traverse a tree anyways and a solid answer on why this isn't a strong development idea for lets say... a game engine, will help cement my understanding on the use of heirarc. rep.
A way to think about this is puppets! Say we want the puppet to raise its hand. We just pull the string associated with that arm, and the hand raises. We didn't have to think about each individual joint and body part involved in raising the hand. Here, the linear representation would require us to consider every part on its own, aka think about every joint and body part, whereas in the hierarchical representation, we just lift the hand and everything else (arm, elbow, forearm) raises itself.
is it appropriate to think of a parent group as defining a coordinate system, or planes, or some kind of space that the subgroups apply themselves to? So when the parent transforms, the subgroups just apply their transforms in the new coordinate system?
When applying transformations to leaves in the hierarchy tree, is there a way of limiting transformations such that they make sense in relation to the parent nodes of the tree? Or would we have to be conscious of that when applying transformations ourselves. For instance, it would be awkward if we made the hand on an arm accidentally slide 10 feet away from the arm it was supposed to be attached to.
In practice how many groups or trees do models have and when calculating the "end-state" of the tree are there optimizations for caching / speeding up calculations of relative coordinates for each step?
So is hierarchical representation relational in the sense that groups are defined with transforms relative to another group (and is that why for example if you raise the hand it would raise the arm--because the arm is defined as a transform relative to the position of the hand?)
@sebzhao I think this is indeed the case. Each new leaf in the hierarchy is defined by features in it's local frame, as well as a transform from the parent frame to the local frame.