We are about to switch to a new forum software. Until then we have removed the registration on this forum.
Hi everyone. I would like use the pointclouds from kinect for link them as mesh, or particular, or other fluid simulation. What is the best way about generate it in real-time and easy to do? GLS shaders or some library as _he_mesh, toxilibs? I do not know programming gls shaders and could help me some editor and after export the code. One exists ?
I can't find a way. Thanx!!
guys do not be shine. I think is a topic usual for lot people...
Using a library such as Hemesh or Toxiclibs is easier than GLSL. Both these libraries are able to create isosurfaces from 3D points. So that seems like a good starting point, especially if you don't know GLSL. Toxiclibs is a little faster than Hemesh. A custom solution would be even faster. Of course in the end, GLSL is the fastest, but also much more complex.
Thank you amnon! So you mean take all pointclouds and create an isomesh from them by toxilibs? I saw toxilib is update until 2011 or am I wrong?
Is it easy link this points by toxilibs? I never used this library still. Some tutorial or doucumentation?
GLSL i would like to learn, I am reading some toutrial, but I am scared 'cause I will need lot time that I haven't.
Do you know if is there some UI for create GLSL shaders and after export the code? It will be perfect :)
It's true that Toxiclibs hasn't been updated in a while. But it may still be useful. I understand you are looking for easy looking solutions. The bad news, there are no easy solutions. ;-) Either try stuff, learn, investigate or just give up now. It's up to you...
I am looking some inspiration from your toutorial on CA. It's look like he_mesh is more update than toxilibs, is it true? And given an array of points, is it easy connect them in mesh and modifiy the mesh in different mode, what do you think?
Some example is always helpful.
" Either try stuff, learn, investigate or just give up now." I am doing it, but I am in a crossroads where I do not know if learn a library, a GLSL code, a Processing Shaders...a lot ways and all not so easy.
The help now maybe is on find the right way and than learn it ;)
Hemesh is still actively developed, Toxiclibs is not. Both libraries contain examples, so you don't need to learn them to try this. You can just install the libary and look at the examples.
Ok, thank you amnon! I saw that you are a guru of shaders and this library :) I found that you create a GUI for he-mesh, but it doesn't work on my Processing. Maybe I wrong installing or do you have any suggest?
He-mesh look very powerful!!
PS why this libraries are slowly than GLSL shaders?
I did once create HemeshGui, but that was a long time ago. So now it doesn't work anymore because of different versions of Processing and libraries. Still, Hemesh is indeed a great library with a lot of possibilities.
These libraries work on the CPU, while shaders work on the GPU and use parallel processing. That's why shaders are much, much faster.
It look like the past time was better than now :) I didn't know that this library use only the cpu!! Hope one day will be a library for GPU.
I read a lot of GLSL this days, but look very hard for me that I came from Media Art, I am software user more than programmer. The language look very rich and difficult to learn...Maybe it is only my sensation, but I am scared that using pointclouds and mesh only by cpu, it will be so slow, even if I have good video card...
PS instead p3d and the beginShape() etc are using always CPU or GPU? Or if I write size(,, OPENGL); it improve the performance or doesn't change nothing?
Your analysis is probably correct. Libraries are easier, but most likely too slow for realtime. GLSL is fast enough, but complex to achieve.
The Processing OpenGL renderers (P2D and P3D) use a lot of GPU to increase the speed, especially PShape and PShader. These will indeed be faster for sketches with a lot of geometry. You could even store geometry created with Hemesh in a PShape for fast display on the GPU. The problem is that in your case, the geometry changes every frame, so you would be constantly creating and having CPU-GPU interaction, thus slow down.
It could be a good choose for you link every points as TRIANGLE or TRIANGLE_STRIP? Which limits I have on this way and which impove? Or better, in which case is better use he-mesh and in which case Pshape or Pshader? All for semply the life but with a performance for real-time.
PS thank you for the time you are spending on this conversation, and I think I am speacking with one of the guru of Shaders, GLSL of Processing. So have if my questions seem banal or stupid is because this topic is not easy to understand by books or reading just the tutorials.
is there any code for generate a 3d pointclouds as array where I can test different way for create mesh/shaders?
Until now I saw he-mesh as work and is impressive. But I would like create one algoritmo for give rules on the connecion points (vertex), and move these points with some noise. In that way i should use only the PShape and vertex, so it will be faster than use he-mesh or he-mesh+Pshape right? Using only PShape i should write P3D on setup () and not OPENGL?
And using PShape can i buffer the frames before and give them some fisics? Hope in an answer
Anyone solved this?