Page 5 of 11

Re: Tuesday, June 27, 2017

Posted: Wed Jul 12, 2017 2:29 pm
by Hyperion
AdamByrd wrote:
Wed Jul 12, 2017 9:11 am
I'm currently digging through isosurface extraction techniques so we can get better meshes out of the procedurally generated objects
So are we talking better meshes as in

A
Spoiler:      SHOW
Image
B
Spoiler:      SHOW
Image
C
Spoiler:      SHOW
Image
D
Spoiler:      SHOW
Image
E
Spoiler:      SHOW
Image
F
Spoiler:      SHOW
Image
G
Spoiler:      SHOW
Image
H
Spoiler:      SHOW
Image
I
Spoiler:      SHOW
Image
J
Spoiler:      SHOW
Image
or a combination of, or ideally all of the above?

Re: Tuesday, June 27, 2017

Posted: Wed Jul 12, 2017 2:35 pm
by Silverware
Not models Hype.
Meshes.

He is trying to get the generation algo to spit out cleaner models so that it'll be easier to manipulate or store/render them without artifacts.
I assume Adam will eventually work on the models too. :V

Re: Tuesday, June 27, 2017

Posted: Wed Jul 12, 2017 2:42 pm
by Hyperion
Aye, but I was more wondering about the flexibility of the meshes for creating interesting and complex geometries and how they might integrate with one another to create smooth models

Re: Tuesday, June 27, 2017

Posted: Wed Jul 12, 2017 8:28 pm
by Flatfingers
DreamKeys wrote:
Wed Jul 12, 2017 10:44 am
I find this array start index thing especially funny, because lua starts arrays by default at index 1 :D
<groan>

That will be me making lots of obi-wan errors, then.

Re: Tuesday, June 27, 2017

Posted: Thu Jul 13, 2017 3:52 am
by Damocles
Interestingly, that one is probably the most plausible/realistic space-vessel in the collection.
Image

Re: Tuesday, June 27, 2017

Posted: Thu Jul 13, 2017 9:15 am
by AdamByrd
Basmannen wrote: Oh, I meant more like, as a novice, where do I start reading to learn about things like that in general? What are the subjects to learn?
By "things like that" do you mean things like surface extraction or just things that are difficult/cool?

Hyperion wrote: So are we talking better meshes as in
< ... >
Silverware is right: meshes not models. Models, when they are generated, are stored in whichever way was most amenable to creation. That may be a binary tree built up through CSG operations or an adaptively sampled distance field in an octree. At the end of the day, however, we need to render that object on a modern GPU which means we need triangles (I mean, there are esoteric methods, but let's stick with the tried-and-true). There are various methods for extracting a surface from these other representations. So we're talking better meshes as in going from the bottom right to top right:

Image

Re: Tuesday, June 27, 2017

Posted: Thu Jul 13, 2017 9:41 am
by Damocles
So the asteroids are procedural volumetric "voxelfields" or similar, and then their surface is then generated by using maching cubes?


here a nice visualization of marching "cubes" in 2D

Image

https://www.volume-gfx.com/

Re: Tuesday, June 27, 2017

Posted: Thu Jul 13, 2017 11:18 am
by Hyperion
Ah, okay. I misunderstood given that I tend to know the term mesh to apply to what you call model. :shifty:
So basically you're having fewer triangles represent the same area, aka less math, aka faster & less likely to bork?

Re: Tuesday, June 27, 2017

Posted: Thu Jul 13, 2017 11:52 am
by Basmannen
AdamByrd wrote:
Thu Jul 13, 2017 9:15 am
By "things like that" do you mean things like surface extraction or just things that are difficult/cool?
I guess what I want to know is where to find resources on things like building a (good) physics engine. Like, I can read about octrees or BSPs but only if I know they exist. I just don't know how to start.

Re: Tuesday, June 27, 2017

Posted: Thu Jul 13, 2017 12:25 pm
by AdamByrd
I prefer textbooks myself. I tend to grab one on a subject and just go cover to cover, supplementing it with blogs and whitepapers I find from general internet searches as necessary. I'm not super up-to-date on the best resources, but we have Realtime Collision Detection and Game Physics Engine Development here in the office. I used RCD for the BSP trees and it was reasonably good. It has some quirks I don't like, but it got the job done.

Re: Tuesday, June 27, 2017

Posted: Thu Jul 13, 2017 12:37 pm
by Basmannen
Thanks for the answer!

Re: Tuesday, June 27, 2017

Posted: Fri Jul 14, 2017 5:34 am
by JFSOCC
Hey Josh, it's been a couple of weeks again. Fight the temptation to wait until you got something big to show us and just post an update.
How's the scheduling btw? Sleep cycle, work cycle, all that?

Re: Tuesday, June 27, 2017

Posted: Fri Jul 14, 2017 5:36 am
by Cornflakes_91
JFSOCC wrote:
Fri Jul 14, 2017 5:34 am
Hey Josh, it's been a couple of weeks again. Fight the temptation to wait until you got something big to show us and just post an update.
How's the scheduling btw? Sleep cycle, work cycle, all that?
AdamByrd wrote: But yea, Josh is going to be slammed until the end of next week.

Re: Tuesday, June 27, 2017

Posted: Sat Jul 15, 2017 2:44 pm
by jonathanredden
I see you guys are taking quite awhile (no rush take all the time you need)

Re: Tuesday, June 27, 2017

Posted: Sun Jul 16, 2017 9:07 am
by Talvieno
Josh is currently swamped with teaching a coding class for local high-schoolers, but he's okay and he'll be back. :)