I used grade_hw2.rib with some varied parameters for grading. The following two images come from that file: (sorry about the jpg artifacts, but you get the idea).
| 180 fov | 130 fov | pitch = 1.3 |
|---|---|---|
| 
| 
|
The most common mistake people made was to not generate circular hex lenses, but rather to make the top and bottom flat as in the following image:
I allowed you to shoot the rays backwards from the eye, since that's what I did for the sample images.
Perhaps a picture will help illustrate.
The fov is centered at the normal. Basically, the fov gives us a way to figure out which rays that fall under the lens don't need to get shot into the environment. You will, unfortunately, still need to shoot them out orthographically (see below for explanation). Assume the fov is rotationally symmetric for the flyseye lens system (ie after applying fov cutoff, you will be left with a circle of rays that strike the lens).
No. For this assignment, we are attempting to model a physical lens system. As such, you do not have to worry about negative fov, fov being greater than 180 degrees, negative pitch, pitch being smaller than 2*radius, or negative radius.
Ideally, you would not have to trace them at all. However, the current build for lrt dosen't let you prohibit the tracing of a ray without mucking with camera.cc. Instead, it will suffice to return an orthographic ray.