First of all, hope title doesn't falsificate and sorry if my question makes few to zero sense.


As I am working on a browser based game with isometric projection, my initial idea was using sprites with canvas, but after then decided to switch to a webGL based solution for using some 3D features as well.

After then and reading Ocean Quigley's blog post about imposters ( http://oceanquigley.blogspot.com/2010/03/decorating-buildings-in-simcity4.html ) , I was not sure about " Buildings were rendered by projecting textures onto very simple 3D geometry " part.

So I sent him an email about this and he was kind to reply. If I may quote he replied that

... we made simple boxes to capture the extents of the buildings. We UV'd the building sprites onto those boxes, using an orthographic projection, mapped to the appropriate camera (which was fixed per zoom). That was just part of our content pipeline - the game fetched the appropriate art asset (3d model with texture) and inserted it into the scene.

After this happy coincidence to learn briefly what UV mapping is, I am pity still not sure about this basic geometry part

For example when looking at Simcity 4's Dao Chemicals building,





If I am not mistaken, entire building without container props next to is one box, container props next to it are separate boxes and game fetches them and put into scene. And doubt these are 3D models anymore looking at Last Chance Motel for example (tables, trashcans, signs all are separate props)



But what I don't get is (considering what I know about UV mapping is very limited) if it is possible to UV map a box with a "texture" like this?

Or asking same question different way, how can I manage mapping sophisticated modelling into simple geometry?


Thanks in advance

It's pretty much exactly as it looks in the article, though I admit they could have chosen better screenshots.

They rendered the high-quality models from a fixed view angle. However, because they needed objects to interact with the Z-buffer properly, they can't use a single billboard; so instead, they use a basic 3D model that gets the rough shape of the building, and project the 2D "sprite" onto that geometry.

It's pretty much exactly as it looks in the article, though I admit they could have chosen better screenshots.

They rendered the high-quality models from a fixed view angle. However, because they needed objects to interact with the Z-buffer properly, they can't use a single billboard; so instead, they use a basic 3D model that gets the rough shape of the building, and project the 2D "sprite" onto that geometry.

I can understand simple geometry for things like billboard but can't get UVing complex buildings with simple geometry part.

So if a projection to rough geometry involves, is there any sophisticated UV mapping in fashion of

or it's more like a rough geometry to serve as bounding box and a sprite?

I cant fully understand what it is that you are asking, so if I do not answer it, let me know.

As per the article, it looks like its just mapping textures onto simple geometry, nothing more to it, and no complex geometry involved.

Offline, You simply turn your hi-res model into a texture based off of the angle you want and add that to a plane. Most 3D modeling software allows you to take a "screenshot" of the model in the viewport. Do this for each face of the primitive, adjusting to make sure it looks right.

What I am trying to ask is (if we put article aside for a moment),

for example at "Dao chemicals" example, isn't it just one bounding box and a sprite? I don't see the part of "we UV'd building sprite to boxes" , isn't it quite complex to be mapped to simple geometry ?

Unduli:

http://imgur.com/a/uAF64

Unduli:

http://imgur.com/a/uAF64

Thank you for your detailed explanation and caring to prepare this all.