0

I am interested in creating a 3D visualization of network packets. A few years ago these things sold for tens of thousands of dollars, but now I think that I can hack one together in a few hours using an open source 3D kit.

I've looked around and have found two kids that look good --- one is Panda3D and the other is CrystalSpace.

My requirements are:

  • Fast to learn
  • Able to run from python or C++
  • Able to work with 50,000 polygons. (I want to represent each packet as a little brick in 3D space.)

This visualization doesn't need to run in a browser.

So I'm looking for advice. My questions:

  1. Which is better for my application, Panda3D or CrystalSpace 3D?
  2. Is there another engine that I should be looking at instead?

Thanks.

RustyTheBoyRobot
  • 5,891
  • 4
  • 36
  • 55
vy32
  • 28,461
  • 37
  • 122
  • 246
  • 1
    A few hours? Either you're some kind of superhuman, or slightly on the ambitious side :) – MattDavey Jun 29 '11 at 16:01
  • Matt, we're all superhuman. Computers are amazing. Any suggestion on what I should use? – vy32 Jun 29 '11 at 21:14
  • I found this good list of 3D references: http://en.wikibooks.org/wiki/Python_Programming/Game_Programming_in_Python – vy32 Jul 02 '11 at 01:09

1 Answers1

1

If you want to get something going in only a few hours I think your only viable option is Visual Python. It is much faster than Panda3d/Python for large quantities of primitives and has a much easier API. It does not have an option to work from C++ but since it is a very thin wrapper of a C++ back end I don't think you would be able to add too much performance dropping the Python. I can compute and display 8000 lit/shaded rotating boxes at 15fps on my system.

JT.
  • 633
  • 1
  • 8
  • 9
  • Do you have links to benchmarks comparing primitive performance between VP and Panda3D? – Russell Borogove Jan 05 '12 at 23:17
  • Thanks. I spent a few hours and got the Panda3D program to create boxes and stuff, but it was a real pain. – vy32 Jan 06 '12 at 03:23
  • Sorry Russel no links for benchmarks. I have implemented two alternatives to Visual Python with similar API but based on PyOpenGL and Panda3d. Both are plenty fast for low quantities of objects. But for a relative data point 1000 boxes in in the pyOpenGL is 2.5fps the Panda3d version is 5.5fps (and might be able to be optimized but loading the models is really really slow) and Visual Python is 120 fps.At some point I'll publish these these libraries (as alternatives for my Visual+ODE physics library [link](http://www.missioncognition.net/software/visualpyode.html)) but they are not ready yet. – JT. Jan 06 '12 at 18:33