Hey first post, go easy!
I’ve just finished finished using jMonkey for a school project - creating buildings at runtime. First of all a big thank you to the jMonkey crew for a great engine, that really sped development of the project. It got used for realtime previews of output and creating a nice interface system. After I got around the rendering-two-swing-windows-at-once thing it worked flawlessly!
created in 10 sec:
Maya mental ray (non real time) renders can be found at http://twak.blogspot.com
It can be downloaded from http://www.cs.bris.ac.uk/home/tk1748/sity.html (sorry, you need the evil java3D caus I was too lazy to port the libraries, dont worry it isn’t really used)
Its real buggy at the moment, and has many glitches, but the ideas there… do you think there is any benefit/demand in working to get this up to the kind of quality where it might be useful?
[edit:] Description: It uses a weighted straight skeleton and Voronoi tessellation as parts of a grammar to create random cities in a couple of seconds. You enter a bunch of parameters and it creates a 300m square section of city for you (or crashes in a horrible mess…)
A straight skeleton is the shape made by the edges of a normal slopey house roof. By changing the slope of different sections to different amounts (inwards or outwards) at different heights you can make a pretty good imitation of a house. A Voronoi tessellation is just a way of subdiving a shape from a set of input points so that each point owns a ‘fair’ amount of space around it. Using these two things in combination produces a really good set of building blocks for creating a city on the fly.
Think of it as aspiring to be a http://www.speedtree.com/ for cities.
twak said:
...
do you think there is any benefit/demand in working to get this up to the kind of quality where it might be useful?
...
what do you intend to use it for?
a quick summary of what it actually does would be nice
just quickly looked through your thesis, and it looks very very interesting! nice work!
this line bothered me a bit though 
Large dependency tree
The concept is definitely very intresting… As a programmer with art skills below 0 I can only dream of procedural content like this 
Indeed, imagine a game like GTA with huge generated cities…Technically speaking, aside from the obvious lack of texturing, I think the 2 features I miss most are the concept of main roads (you mention this in the paper) and lower/smaller buildings/areas.
I can only say that if you package up your work to make it a bit more easy to use (like no unnecessary dependencies such as installing Java3D, blegh), it will more easily get the attention it deserves.
Did you make the buisness plan purely for the school assigment?
mr coder got me curious and i had to (quick)read the bussiness plan and the thesis. and also mr. coder is right: jme has no dependency on jogl at all.
the idea is very interesting. i didn't find anything about texturing, but i suppose you already have some ideas queued for implementation 
well, as for your previous question: the biggest problem is see is (as you stated in the business plan) "Product only applicable to games with urban content". you marked the risk factor as "low", but i see that a bit different. sure, urban content is required, but the amount of such required content has to be big enough to require the use of a tool like sity. imho there are not enough titles/year which will use that much urban content.
oh, i found this part very amusing btw: "…It is intended as a stand alone real time solution to entire world generation, but has not seen commercial use or been updated since 2003. The price for this engine is estimated at
Hey guys! thanks for the replies, I'm impressed you bothered reading it!
Yah, the business plan is totally unworkable. My cost/benefit showed that it was, but it was for a school project so (with a little help from goal-seek in excel) the figures where massaged to show it would work. It's the kind of project that could only come from the open source (or academic) world! I think you'd have to do it because you loved it as theres some serious work involved in just beginning to make it useful. But I can see lots of ways of getting the work published if it flies!
Do'h I wasn't expecting people to really read that paper 
I didnt quite know what i was talking about I assumed (wrongly) that jME used LWJGL that used JOGL. But I guess not.
Pascal M
Looks quite nice!
I’m printing out your thesis at the moment, and will read through it later.
I’m interested in city generation myself too - I just finished my thesis on real time landscape generation (http://www.iki.fi/zzorn/landscapegeneration). It reviews some existing work about city generation, although I have some ideas I’d like to develop further (e.g. how to generate cities and neighbourhoods with distinct appearances, instead of homogeneous cities). Currently I’m working on terrain generation though.
Are you planning to release the source? Did you find good straight skeleton and Voronoi algorithms on the net, or did you write them yourself? I know I will be needing both of them sooner or later, and I haven’t yet found an O(n log n) performant open source Java Voronoi algorithm on the net.
– zzorn
I'll put my code up when it all gets graded, but it is not a Fortune's n log n Voronoi implementation. Its a lot slower (n*n) and buggier (…well only had 10 weeks to implement a lot!). Its on my list of things to do in java, that and a good straight skeleton imelementation are really needed in the language…
zzorn your thesis wont render on xpdf or acrobat!
Okay, thanks anyway.
If I get around to writing an O(n log n) Voronoi algorithm in Java I'll upload it on sourceforge then and drop a note here also.
Thanks for pointing out the rendering problem, perhaps I'll need to include the fonts? I'll look into it.
It works for me in Windows XP and Adobe reader 7.0. Are you using Linux?
– zzorn
Zzorn! - i've just tried loading your thesis on my pc and it renders fine in windows, but doesnt like any readers of mac/ os x! I guess theres something wrong with the fonts… Great overview of content generation tho - you covered so much!
I was just looking at the videos of Alan Wake
http://www.youtube.com/watch?v=CLEBOn9K5Nk
Imagine being able to use a whole core to generate content on the fly, another to run AI and a final one to run physics!
Thanks!
Did you try re-downloading it recently for the mac? There seemed to be a problem with hosting it on googlepages, for me the download did not always finish when I tried, leading to corrupted pdf files. I moved the thesis to another site yesterday ( http://purple.worldforge.org/~zzorn/thesis/LandscapeGeneration.pdf ).
The multi core processor demo was impressive indeed, especially the volumetric lighting.
– zzorn
twak said:
Zzorn! - i've just tried loading your thesis on my pc and it renders fine in windows, but doesnt like any readers of mac/ os x! I guess theres something wrong with the fonts... Great overview of content generation tho - you covered so much!
I was just looking at the videos of Alan Wake
http://www.youtube.com/watch?v=CLEBOn9K5Nk
Imagine being able to use a whole core to generate content on the fly, another to run AI and a final one to run physics!
i spoke with their production lead(lasse sepp