Lecture 9: Ray Tracing & Acceleration Structures (41)
colinsteidtmann
What are the "objects" here? It looks like infinite objects almost. Is every water drop an object, or every grass blade?
srikartalluri
In my understanding. I do believe that every every grass blade would be a separate object, especially on the left side. On the right side, it becomes a little simpler as we have leaves for each of the planets. However, the water I think is separated into bigger chunks representing the surface of the stream rather than each individual droplet.
AlsonC
How does the grid work when certain plants stick out, for example on the right side, taller plants seem to take more grids than for example the smaller leaves below or the grass on the left.
Zackoon
To add on, how would we determine the grid resolution here? If we must pre-determine it via C * # objects, and a blade of grass = 1 object, where do we get the overall number? Is it just an estimate usually?
AbhiAlderman
How specific is our definition of objects? Is it just based on size taken up on the screen? For example, I imagine that the leaves on a tree would be classified as different objects as long as the tree was close enough. But would the irregular shape of the bark on the tree be classified as different objects, since some areas may protrude more than others? If instead we were looking at a car, would the tires, mirrors, doors, and windows be different objects? Or depending on the distance/scale of the car, would the entire car be classified as one object? Are objects only separated by the different models available in the creation of the image? Is there some sort of algorithm or method for a computer to make these decisions, or is it based entirely on the actual created objects in a scene and the creator's discretion?
weinatalie
In applications such as basic game development, the approximate number of squares in the grid are determined based on the number of collide-able objects. For a densely populated scene like the one shown, it makes sense to have a uniform grid with as many squares as there are objects; as all the objects are evenly distributed, there is likely to be a collision in each square. For another scene that is less evenly populated, it would not make sense to have a grid with as many cells as there are objects. There would be no collisions in a large portion of the grid, making it computationally quite wasteful.
What are the "objects" here? It looks like infinite objects almost. Is every water drop an object, or every grass blade?
In my understanding. I do believe that every every grass blade would be a separate object, especially on the left side. On the right side, it becomes a little simpler as we have leaves for each of the planets. However, the water I think is separated into bigger chunks representing the surface of the stream rather than each individual droplet.
How does the grid work when certain plants stick out, for example on the right side, taller plants seem to take more grids than for example the smaller leaves below or the grass on the left.
To add on, how would we determine the grid resolution here? If we must pre-determine it via C * # objects, and a blade of grass = 1 object, where do we get the overall number? Is it just an estimate usually?
How specific is our definition of objects? Is it just based on size taken up on the screen? For example, I imagine that the leaves on a tree would be classified as different objects as long as the tree was close enough. But would the irregular shape of the bark on the tree be classified as different objects, since some areas may protrude more than others? If instead we were looking at a car, would the tires, mirrors, doors, and windows be different objects? Or depending on the distance/scale of the car, would the entire car be classified as one object? Are objects only separated by the different models available in the creation of the image? Is there some sort of algorithm or method for a computer to make these decisions, or is it based entirely on the actual created objects in a scene and the creator's discretion?
In applications such as basic game development, the approximate number of squares in the grid are determined based on the number of collide-able objects. For a densely populated scene like the one shown, it makes sense to have a uniform grid with as many squares as there are objects; as all the objects are evenly distributed, there is likely to be a collision in each square. For another scene that is less evenly populated, it would not make sense to have a grid with as many cells as there are objects. There would be no collisions in a large portion of the grid, making it computationally quite wasteful.