I'm still on course on my huge Phoenix photoreal project (I keep finding new really cool areas to add to the peripherals) and have found the most daunting aspect to the whole process is laying down decent autogen buildings and vegetation on the over thousands of scenery tiles that have been created. Annotator from the Microsoft Scenery SDK is a basic program but it's giving me carpal tunnel having to draw the box around every building and vegetation area. Some of the tiles are huge housing developments and to get the real effect of the myriad of homes (especially the night lighting from them) you have to create hundreds of little tiny buildings per tile.
Here's what I'd imagine a great autogen creation program to be. First of all, the window in which you choose which tile you want to edit clearly shows which ones you've already done and possibly a thumbnail rendering of the image and the density of the autogen on it.
Secondly, I don't want to have to define for each file I edit the building heights or the vegetation distribution. Annotator lets you default to the last used values but that requires a great amount of mouse movement and clicking before you even begin to edit. If you don't define them, you end up with huge buildings. An area like Phoenix, the buildings rarely get up to even 2 stories.
Now this would be the ultimate. Instead of going in and drawing the little boxes where you want the buildings, why can't you just point to the middle of the building in the image and some basic AI computes the corners for you. Most of the industrial buildings I put in have very white roofs. Should be easy to write code to look for the extents of the light colored pixels and estimate the box. If it does a poor job you have the ability to override the choices.
You should be able to define an area as a housing development and specify the house size. The program then goes in and, based on the grid it finds in that part of the image, randomly puts in houses for you.
An interface to FS would enable you to slew over the area you wanted autogen in and the program would pick out the scenery tile you needed to edit.
Man, what a time saver this would be. I might just tackle this one myself. Need a good excuse to get proficient on VB.net. Anyone had any experience with programming that queries image pixels? Is there a routine already written out there for taking a DXT1 file and converting it back into a standard bitmap image?
Art Martin
Here's what I'd imagine a great autogen creation program to be. First of all, the window in which you choose which tile you want to edit clearly shows which ones you've already done and possibly a thumbnail rendering of the image and the density of the autogen on it.
Secondly, I don't want to have to define for each file I edit the building heights or the vegetation distribution. Annotator lets you default to the last used values but that requires a great amount of mouse movement and clicking before you even begin to edit. If you don't define them, you end up with huge buildings. An area like Phoenix, the buildings rarely get up to even 2 stories.
Now this would be the ultimate. Instead of going in and drawing the little boxes where you want the buildings, why can't you just point to the middle of the building in the image and some basic AI computes the corners for you. Most of the industrial buildings I put in have very white roofs. Should be easy to write code to look for the extents of the light colored pixels and estimate the box. If it does a poor job you have the ability to override the choices.
You should be able to define an area as a housing development and specify the house size. The program then goes in and, based on the grid it finds in that part of the image, randomly puts in houses for you.
An interface to FS would enable you to slew over the area you wanted autogen in and the program would pick out the scenery tile you needed to edit.
Man, what a time saver this would be. I might just tackle this one myself. Need a good excuse to get proficient on VB.net. Anyone had any experience with programming that queries image pixels? Is there a routine already written out there for taking a DXT1 file and converting it back into a standard bitmap image?
Art Martin
.
