|
|
Post by Jamis on Nov 21, 2018 17:23:49 GMT
Here's the scene description for the image at the start of chapter 11, "Reflection and Refraction". It renders the following picture:  The scene is described in YAML format, which is hopefully self-explanatory, but if something doesn't make sense, please let me know.
# ======================================================
# reflect-refract.yml
#
# This file describes the scene illustrated at the start
# of chapter 11, "Reflection and Refraction", in "The
# Ray Tracer Challenge"
#
# by Jamis Buck <[email protected]>
# ======================================================
# ======================================================
# the camera
# ======================================================
- add: camera
width: 400
height: 200
field-of-view: 1.152
from: [-2.6, 1.5, -3.9]
to: [-0.6, 1, -0.8]
up: [0, 1, 0]
# ======================================================
# light sources
# ======================================================
- add: light
at: [-4.9, 4.9, -1]
intensity: [1, 1, 1]
# ======================================================
# define constants to avoid duplication
# ======================================================
- define: wall-material
value:
pattern:
type: stripes
colors:
- [0.45, 0.45, 0.45]
- [0.55, 0.55, 0.55]
transform:
- [ scale, 0.25, 0.25, 0.25 ]
- [ rotate-y, 1.5708 ]
ambient: 0
diffuse: 0.4
specular: 0
reflective: 0.3
# ======================================================
# describe the elements of the scene
# ======================================================
# the checkered floor
- add: plane
transform:
- [ rotate-y, 0.31415 ]
material:
pattern:
type: checkers
colors:
- [0.35, 0.35, 0.35]
- [0.65, 0.65, 0.65]
specular: 0
reflective: 0.4
# the ceiling
- add: plane
transform:
- [ translate, 0, 5, 0 ]
material:
color: [0.8, 0.8, 0.8]
ambient: 0.3
specular: 0
# west wall
- add: plane
transform:
- [ rotate-y, 1.5708 ] # orient texture
- [ rotate-z, 1.5708 ] # rotate to vertical
- [ translate, -5, 0, 0 ]
material: wall-material
# east wall
- add: plane
transform:
- [ rotate-y, 1.5708 ] # orient texture
- [ rotate-z, 1.5708 ] # rotate to vertical
- [ translate, 5, 0, 0 ]
material: wall-material
# north wall
- add: plane
transform:
- [ rotate-x, 1.5708 ] # rotate to vertical
- [ translate, 0, 0, 5 ]
material: wall-material
# south wall
- add: plane
transform:
- [ rotate-x, 1.5708 ] # rotate to vertical
- [ translate, 0, 0, -5 ]
material: wall-material
# ----------------------
# background balls
# ----------------------
- add: sphere
transform:
- [ scale, 0.4, 0.4, 0.4 ]
- [ translate, 4.6, 0.4, 1 ]
material:
color: [0.8, 0.5, 0.3]
shininess: 50
- add: sphere
transform:
- [ scale, 0.3, 0.3, 0.3 ]
- [ translate, 4.7, 0.3, 0.4 ]
material:
color: [0.9, 0.4, 0.5]
shininess: 50
- add: sphere
transform:
- [ scale, 0.5, 0.5, 0.5 ]
- [ translate, -1, 0.5, 4.5 ]
material:
color: [0.4, 0.9, 0.6]
shininess: 50
- add: sphere
transform:
- [ scale, 0.3, 0.3, 0.3 ]
- [ translate, -1.7, 0.3, 4.7 ]
material:
color: [0.4, 0.6, 0.9]
shininess: 50
# ----------------------
# foreground balls
# ----------------------
# red sphere
- add: sphere
transform:
- [ translate, -0.6, 1, 0.6 ]
material:
color: [1, 0.3, 0.2]
specular: 0.4
shininess: 5
# blue glass sphere
- add: sphere
transform:
- [ scale, 0.7, 0.7, 0.7 ]
- [ translate, 0.6, 0.7, -0.6 ]
material:
color: [0, 0, 0.2]
ambient: 0
diffuse: 0.4
specular: 0.9
shininess: 300
reflective: 0.9
transparency: 0.9
refractive-index: 1.5
# green glass sphere
- add: sphere
transform:
- [ scale, 0.5, 0.5, 0.5 ]
- [ translate, -0.7, 0.5, -0.8 ]
material:
color: [0, 0.2, 0]
ambient: 0
diffuse: 0.4
specular: 0.9
shininess: 300
reflective: 0.9
transparency: 0.9
refractive-index: 1.5
|
|
|
|
Post by salver on Dec 20, 2018 21:50:56 GMT
When I follow the description I get a picture that is quite similar to your picture, but the two spheres in the front do not have the correct reflection in them. All tests are green, and I don't understand what is wrong. I attach a screenshot and the ppm file. B5Reflect.ppm (748.1 KB)  P.S. Great book! |
|
|
|
Post by Jamis on Dec 20, 2018 23:32:27 GMT
salver, judging by the picture you shared (thanks for doing that!), my guess is that the spawned rays are not being refracted properly. They appear to be passing directly through, as if they were soap bubbles filled with air. Can you double check that you've set the index of refraction on those spheres? It should be set to 1.5 (to mimic the IOR of glass).
|
|
|
|
Post by salver on Dec 21, 2018 15:37:27 GMT
The refractive-index is set to 1.5 for both spheres.
|
|
|
|
Post by Jamis on Dec 21, 2018 15:48:48 GMT
Interesting, okay. Is your code available anywhere online? If so, I'd be happy to try and troubleshoot a bit. Otherwise, I'd recommend checking out the value of "n1" and "n2" when a ray strikes either of those spheres. You might also check the value of "cos_t" and the direction vector of the refracted ray, in the "refracted_color" function. Right now, it looks (from the picture) like the new direction vector is the same as the direction vector of the incoming ray, and it shouldn't be.
|
|
|
|
Post by salver on Dec 21, 2018 16:43:53 GMT
I will try to investigate the parameters that you pointed out.
If I don't find anything I will upload the code to Github.
|
|
|
|
Post by salver on Dec 22, 2018 16:24:15 GMT
I have problems with the troubleshooting, can't figure out how to stop at the most interesting places, and how to decide if a value is OK or not. If the book had an extra chapter with tips / examples of troubleshooting, that would have been really nice.  I have copied the source to Github: github.com/stein-a/RayTracer Please have a look if you like. But you don't have to do it - I wouldn't like to ruin your Christmas. |
|
|
|
Post by Jamis on Dec 22, 2018 18:56:43 GMT
salver , that's an excellent point about tips and tricks for debugging. It's sadly too late for the book, but maybe I can compile a post or two to discuss it. Here's what I would do in this case: - Identify which pixel in the scene you want to test. I typically render the image once, and then use Preview.app on the Mac to click on the pixel in question and then draw a box back to the top-left of the image. The box will show its size as you drag, so you can see what the coordinates of the pixel are. (The specific method you use to find the coordinates will probably vary depending on what program you use.)
- Once you know the (x, y) coordinates of the pixel, you can edit your camera's "render" function so that it skips rendering any pixel that is not at the given (x, y).
- Then, add debugging information anywhere you need in your program. I'm a huge fan of printf() debugging, but use whatever works for you.
- Once that's all in place, running your program will attempt to render just that one pixel, and you can get at the debugging info you set up.
Thank you for posting your ray tracer code! I'll make some time today to look at it and see if I can spot anything suspicious. - Jamis
|
|
fs
New Member
Posts: 28
|
Post by fs on Dec 22, 2018 19:19:57 GMT
|
|
|
|
Post by salver on Dec 22, 2018 21:41:46 GMT
Thank you, Jamis, for that nice recipe for troubleshooting. I tried adding printf, without limiting the render, and that gave a LOT of output... Thank you, fs, for pointing out that bug. It was stupid of me to make xs an optional parameter. (I think it was to avoid having to change some of the tests..) I have changed the code and started a new run. Hopefully I will get a perfect picture this time. Updated:  |
|
|
|
Post by ascotti on Dec 26, 2018 19:14:10 GMT
Finally got mine as well, had to code it manually as I don't have a YAML importer yet. Took slightly over 4 minutes at 64 rays per pixel, one of the longest renders I've done so far!  |
|
|
|
Post by thegastropod on Jan 5, 2019 22:18:43 GMT
Hah! I'm kind of running into the opposite problem as salver. My transparent objects always come out WAY too dark. Following the foreground parts of your scene description, mine looks like this. I *suspect* it has something to do with my schlick formula ( github.com/leeourand/romano/blob/master/lib/romano/intersection.ex#L81-L101) but I can't quite figure out what's going wrong. My tests are all green, and the n1, n2 calculations seem right, too. Any ideas what I might be goofin' up? Attachments:
|
|
|
|
Post by ascotti on Jan 5, 2019 23:18:30 GMT
It looks like the transparency level is very low. Not sure it's the reflectance formula, because if that formula brings the transparency down then reflectivity should become high and it doesn't show in the image. But maybe have it return 0 and see what happens. I think an experiment worth doing is multiplying the refracted color by some amount and see if it makes a difference in the image.
|
|
|
|
Post by thegastropod on Jan 6, 2019 0:38:25 GMT
Ahh, good point! So if I set reflectance to 1, I get perfect mirrors. If I set it to 0, I get almost totally black spheres. So you're right. The problem's *not* with that function. Something's not right about my refracted_color value. I'll keep poking around. Thanks for the pointerAttachments:
|
|
fs
New Member
Posts: 28
|
Post by fs on Jan 6, 2019 23:52:39 GMT
I tried a hack to force n1, n2 for the rays transitioning from index 1.5 -> 1.0 (rays exiting the foreground spheres) and it looks like i could reproduce your case:  In my equivalent of determine_n1_and_n2, I included this hack to simulate incorrectly determining n1 to be infinity (the foreground spheres have ref idx 1.5 and the "air" has ref idx 1.0): if n1 == 1.5 && n2 == 1.0 {
n1 = PosInf
}
Maybe it's worth re-checking the n1, n2 that refracted rays exiting those spheres produce, or to try special-casing the n1,n2 results produced by that function so you can see if that might give you a lead. [ I haven't used Elixir, but was wondering if github.com/leeourand/romano/blob/master/lib/romano/intersection.ex#L54 should be using the same approach as github.com/leeourand/romano/blob/master/lib/romano/intersection.ex#L68 to optionally update n1 only if intersection == hit ? ] |
|
