Lecture 9: Ray Tracing & Acceleration Structures (53)
srikartalluri
This is very similar to Binary search. However, what I do not understand is the Termination criteria. It is relatively simple to implement a maximum tree depth however, I am not sure how we can compute the expect cost of ray intersection. IN my understanding, at any given tree node, we would need to actually do the computations to understand how costly it could be.
joyjwlee
Along the same line as @srikartalluri's comments, where do the values 8 and 1.3 come from? Are those values obtained through experimentation?
misha-wu
how does the split plane compare to each choice? can a render use both options selectively?
sebzhao
I think empirically means that from experiments or people's experiences, that's a good number to use. Since we can't know the expected cost of ray intersection if we split, I think the empirical results are a proxy to use instead.
This is very similar to Binary search. However, what I do not understand is the Termination criteria. It is relatively simple to implement a maximum tree depth however, I am not sure how we can compute the expect cost of ray intersection. IN my understanding, at any given tree node, we would need to actually do the computations to understand how costly it could be.
Along the same line as @srikartalluri's comments, where do the values 8 and 1.3 come from? Are those values obtained through experimentation?
how does the split plane compare to each choice? can a render use both options selectively?
I think empirically means that from experiments or people's experiences, that's a good number to use. Since we can't know the expected cost of ray intersection if we split, I think the empirical results are a proxy to use instead.