Functional noob here, just found Irrlicht and look forward to learning how it's put together. I'm using VS 2003 as my IDE, and am working to get acquainted with C++/C# programming after a long time away.
Would appreciate good resources on the most effective tools for bouncing light off of objects. My theoretical interest is reversing those tools to try to take video and identify object boundaries. I know little of what's being done in the Open source community on this subject.
My assumption is that Fourier transforms might be part of a decent route, and I'm comfortable with a some degree of vector calculus.
Also, I'm wondering what data structures people might recommend for storing 3D objects. How does Irrlicht handle this? Has anyone found or developed SQL-based structures?
Thanks for any help. It's probably obvious that I groping around in the dark, even though I feel like I have some vague idea of where to look.
Object Boundary Detection
-
pfo
- Posts: 370
- Joined: Mon Aug 29, 2005 10:54 pm
- Location: http://web.utk.edu/~pfox1
Really depends on what the object is, Irrlicht stores things in mesh buffers (collections of vertices and indices w/ texture coords), and each mesh buffer has 1 material. However, depending on what type of node you use it handles them differently internally (IAnimatedMeshSceneNode is handled different from ITerrainSceneNode, and IOCTTreeSceneNode). But the underlying data structures in each are the same.I'm wondering what data structures people might recommend for storing 3D objects.
Ehhhh.. that might get kinda problematic. A database is going to be slower than regular file I/O, period. If the database were really small, there shouldn't be a problem, but I don't see any point in using SQL-based data over regular file data unless you plan to make heavy use of SQL features, which of course can lead to more slowdown.Has anyone found or developed SQL-based structures?
Most effective is kinda misleading. Since this is a graphics engine, most effective for us means fastest while still looking OK. For Pixar, most effective would be best looking and they wouldn't care how long it took to render. You could take a look at Irrlicht's lighting code, but I think if you looked at some lightmapping code you'd find a lot more info on how lighting calculations are done (I personally know little on the subject).Would appreciate good resources on the most effective tools for bouncing light off of objects.
I probably don't know what I'm talking about when I say this, but I don't think this will help much. Lighting calculations do their thing on 3d geometry that they know is there, whereas if you have a video you have a sequence of images w/o 3d geometry, how could you perform a reverse lighting calculation? It seems like boundary detection in a video would be more like OCR, with a much broader range of things it has to identify, coupled with the challenge that it has to perceive these boundaries across multiple frame of video and make some sense out of it.My theoretical interest is reversing those tools to try to take video and identify object boundaries.
-
barrydingle
- Posts: 3
- Joined: Wed Nov 23, 2005 6:39 pm
Thanks for the reply. I think the best way for me to wrap my brain around this is going to be lots of communication and trial and error, so thanks for the opportunity to get started.
On data:
I'll spend some time getting familiar with mesh buffers.
On the question of SQL, I'll admit I don't know much at this point. I was thinking less for each and every object, and more for how you might handle a large world that extends beyond the viewable area. The closest I've found on the forum so far is this discussion:
http://irrlicht.sourceforge.net/phpBB2/ ... t=database
I think just understanding this piece and better comparison of file i/o can keep me busy for a while.
On lighting:
I'll check out some lightmapping code. Again, thanks.
As for boundary detection, it at least seems to me that with a lightmapped scene, you know where the boundaries are, and therefore could perhaps use data from those boundaries to populate a statistical frame of reference.
Perhaps the field of edge detection already has plenty on that that I could access somewhere?
http://en.wikipedia.org/wiki/Edge_detection
Are you aware of any place in this forum that deals with Laplacian or Gaussian functions? If I can find the right mathematical tools, maybe I can play around with it on my own.
On data:
I'll spend some time getting familiar with mesh buffers.
On the question of SQL, I'll admit I don't know much at this point. I was thinking less for each and every object, and more for how you might handle a large world that extends beyond the viewable area. The closest I've found on the forum so far is this discussion:
http://irrlicht.sourceforge.net/phpBB2/ ... t=database
I think just understanding this piece and better comparison of file i/o can keep me busy for a while.
On lighting:
I'll check out some lightmapping code. Again, thanks.
As for boundary detection, it at least seems to me that with a lightmapped scene, you know where the boundaries are, and therefore could perhaps use data from those boundaries to populate a statistical frame of reference.
Perhaps the field of edge detection already has plenty on that that I could access somewhere?
http://en.wikipedia.org/wiki/Edge_detection
Are you aware of any place in this forum that deals with Laplacian or Gaussian functions? If I can find the right mathematical tools, maybe I can play around with it on my own.