After a few hours of fiddling with sizes today I’ve finished texturing the character models for the crowd simulation. I’m using Oliver’s original textures as I want to focus on the body-shape changes and how they affect the visual variety right now. Next on my plate are the wonderfully time-consuming process of fixing all the normals… I’d lie if I said I was looking forward to that bit…
It’s not overly noticeable from the photo above as everyone is rather small, but there is most definitely an increase in the variety between the people in the original simulation and the new one! My hope is that I can run a test alongside the ones I have planned which uses extra textures… this will help me develop a conclusion as to whether the geometric variety is less important than the texture variety or not.
So I’ve not really blogged much about what my third year project is actually about, so I’ll do a little post now about it, and how things are developing so far.
My project’s title is “Towards Developing Character Visual Variety in a Crowd Simulation Program” and it’s pretty much that: Developing a way of generating varied characters in a crowd simulation!
When simulating crowds in virtual environments, specifically games – it is often the case that models will be re-used, lowering visual variety in the crowd. This project aims to address the problem of re-used models by developing a procedurally generation system in OpenGL to generate more varied character models. The project will then be evaluated using focus groups and a “spot the same models” activity similar to that of McDonnell et al. (2008).
So I’ve been working towards developing a method of generating characters in real-time based on manipulating a base character using several different visual variation techniques the main being:
- Mesh Deformation based on William Herbert Sheldon‘s Somatotype Theory.
- Randomly generated Heights and Widths.
- Randomly chosen Bodies and Heads.
- Randomly generated colour differences in certain areas of character models.
I’m working on Oliver Szymanezyk’s crowd simulation program that he is developing for his post-graduate research project. He’s been working on creating a more realistic crowd simulation in which members of the crowd interact with each other in groups and not just as individual entities:
I’ve been tasked with making the characters more visually varied in the scene. If you look at the video and then at my screenshot below you can notice the differences I’ve added in so far between the old version and my current implementation. Ignoring the texturing problems on the bodies of course… I’m working on editing the texturing soon – since I’ve added in all these new vertices to the body I’ve got to change the way the body reacts to texturing.
After a great deal of painful fiddling with OpenGL vertex arrays and raging over little buggy rendering issues, I’m finally getting somewhere with this final year project! I’ve got my model (mostly) finished, and I’m happy with the results – might do a few more edits before I sign off on it, but the models are looking a great deal more varied than they originally were.
I hope to have the models completed one-hundred percent by early January so then I can hit my deadline for starting the evaluation process of my project. Then I’ll get on the typing up of my final submission… I’ve got most of my notes collated together, it’s just a matter of pulling it all into one epically sized document!