random hemisphere (tlrcam related

oodmb · Post by **oodmb** » Wed Oct 10, 2007 3:02 pm

right, i have no idea what you mean by unterstud.
if you mean by parimeters set in the file, no they arent done like that yet. right now i have an interface for setting parimeters.

if you mean file format for the scene- there is a .tlr which is basicaly a txt which holds a slightly modified .obj file with some material definitions.

recent picture:

what i did to compensate for the brightness of the caustics in respect to the rest of the scene was to multiply the actual diffuse colour by 1/pi and then by 1/pi again for the direct light calculation. in this image the diffuse colour for the walls were .02 .02 .4 .4 .4 .4 .4 .02 .03 the emiter was scaled way up

oodmb · Post by **oodmb** » Fri Oct 12, 2007 3:11 pm

as it turned out, that was not the issue at all. i was wasting my time playing with the path tracing classes as the problem was in the metropolis class and in how Fresnel worked.

its now fixed and in theory a bit faster.
i might release the source soon, apparently if i copywriter it or something along those lines before i post the source online it will actually be an easier way to send to colleges. i won't be able to put it on source forge till after i apply to college.

i want to see if i cant add bidirectional and scattering in a month.

Zom-B · Post by **Zom-B** » Fri Oct 12, 2007 3:23 pm

oodmb wrote:i might release the source soon, apparently if i copywriter it or something along those lines before i post the source online it will actually be an easier way to send to colleges.

Be aware of Behrendt, he's looking for sources for his project

+ a final

oodmb · Post by **oodmb** » Fri Oct 12, 2007 11:48 pm

give it a couple of weeks and i'll see what i have to do, if any body knows the subtleties of this issue, please do speak up.

new pic at 400 mutations, rendered overnight. i dont know why its so grainy, but at least it works. maybe rendering it over the weekend will serve it better.

Kram1032 · Post by **Kram1032** » Sat Oct 13, 2007 9:48 am

looks nice but your radiosity is still too strong...
...and your grain looks like heavy fireflies...
the glass looks quite accurate, though

oodmb · Post by **oodmb** » Mon Oct 15, 2007 5:35 am

well, never said it was perfect. its not realy a profesional renderer anyway, i want to open source it more for people who want to start to write there own renderers rather than for development or for artistic use.

Kram1032 · Post by **Kram1032** » Mon Oct 15, 2007 5:44 am

so to say, you want to make a "starting kit for everyone who's interested"

I'd say, you should at least kill the fire flies

the radiosity actually also is a taste-story

oodmb · Post by **oodmb** » Sun Oct 28, 2007 5:25 am

i fixed the radiosity issue by adding tonemapping. i changed the traversal code to handle more complex things much faster and changed the build code a bit. it now renders things about 6 times as fast. i added/fixed bidirectional. in order to get it working much faster though i need to add ray catching. the only reason i have not done this yet is that i can do it for a pin hole camera, but not yet for a thin lens camera.

IanT · Post by **IanT** » Sun Oct 28, 2007 5:27 am

Wow that looks like excellent progress and nice to see a non-Cornell render

One thing you should be careful of ... tonemapping will only have masked the over-bright GI problem, it won't have fixed it.

Ian.

oodmb · Post by **oodmb** » Sun Oct 28, 2007 5:29 am

i'm pretty sure its also fixed.

random hemisphere (tlrcam related

Who is online