I'm not sure this addresses the question, but the notion of "resetObject"
in to reset the object from the point of view of the network.  When
SuperviseState executes, there are two possible objects for it to output:
one, the input to SS, and the other, the current state of the object as it
has been modified by direct interaction in a UserInteractor.   If
resetObject is 0, then it ignores the object coming into its input and
returns the object that has potentially been modified in a user interactor.
If resetObject is 1, then it uses the input object and discards the changes
made to the previous object by direct interaction.   What all this adds up
to is that if the network makes a change to the object, you should set
resetObject to 1for one execution so that the new object percolates into
the rendered image,, then (if you are modifying it in a user interactor)
reset it to 0.

I always thought that the default here was backward, since modifying the
object in a direct interactor is less common than modifying it in the
network.

Greg

Reply via email to