CSCI 140 Group Gallery

3-D Computer Modeling 
CSCI 140 Spring 2003

Japanese Tea Room

Previous     Next
Alisha Thayer
John Haren
Ross Hertzog
Chris Peterson

Japanese Tea Room
by
The A-Team		Group Scene
 
Inspiration:  
     
Finding a subject matter for this project seemed relatively effortless.  As a group, we probably spent a whole five or so minutes brainstorming before John said, 'Well, I was thinking a Japanese theme would be kinda cool,' at which point everyone agreed, and our subject matter had been chosen.  
     
The decision to make the focus of the project a Japanese Tea Room was also an effortless and painless decision that seemed to pretty much fall into our laps.  Though our initial idea was to make a Japanese garden of some type, the implications that went along with that idea (in regards to modeling, texturing, et cetera) were just overwhelming, so we decided on an inside scene, and for some reason a tea room was the logical choice.
 
 
 
Objects:Moon Window GridRoss
 Shoji ScreensRoss
 TeapotRoss
 CushionsRoss
Katana and StandRoss
 Basic Room ArchitectureJohn
 Rice BowlsJohn
 ChopsticksChris
 GongChris
 Floor LanternsChris
 SandalsChris
 Central TableChris
 TeacupsChris
 VaseChris
 Small WhiskChris
 ArmoireAlisha
 Folding ScreenAlisha
 Hanging Lanterns (modified from Chris’ floor lanterns)Alisha
 Table Under Moon WindowAlisha
 CabinetAlisha
 Paintings on WallsAlisha
 Tatami MatsAlisha
 RaftersAlisha
Textures:woodgood1.jpgChris
 gong texturesChris
 Sandal texturesChris
 Vase/Teacups texturesChris
 TatamiEdited.jpgRoss and John
 WallTile1.jpgRoss and John
 WallTile2.jpgRoss and John
 WallTile3.jpgRoss and John
 Wood41.jpgastronomy.swin.edu.au/~pbourke/ texture/woodgrain
 TeakWoodwww.cc.jyu.fi/atk/blender/blacksmith/ textures/teak.jpg
 IF00004www.allposters.com
 IF00003www.allposters.com
 IF00002www.allposters.com
   
LayoutObject ArrangementAlisha
 LightingAlisha and John
 
Problems and Solutions:
          
Alisha:

Honestly, I had expected more grief from this project that I actually experienced.  Our group worked like a sort of hive mind on this project, which made it go amazingly smoothly, and problems were really few and far between.   Honestly, I spent the most time fiddling with objects in the scene to get them in the absolutely perfect spot relative to other objects, since it was my job to take everyone's objects and arrange them in the scene.  

Though easily, the most frustration  again was with the lighting.  As I had discovered volumetric lighting and radiosity during this project, there were so many things that I had control over now in my lights that it was actually quite overwhelming at times, and rendering test renders became more and more difficult as the number of point and spot lights increased, since render time skyrockets when you have multiple lights with volumetrics enabled (and VIPER, while valuable for testing how your lights work, just isn't the real thing).  However, the biggest lesson I learned during this project is LESS IS MORE!  9 Volumetric lights were actually quite detrimental to the scene, which was quite opposite of what I would have thought.  So, in the end we decided on two actual volumetric lights (the outside light sources) and tweaked the other seven light sources to light up the scene a little more and cast the shadows we were looking for.

Speaking of shadows, that was another horror of this project in regards to lighting was the amount of shadows all our light sources were casting.  We all learned a valuable lesson in what exactly ray-trace shadows does to render times (YUCK!).  

Chris:

The problems I mainly came across were how to create objects really low poly while still making them look as realistic as possible. Because our scene has so many objects I found that every polygon I could eliminate from my model would help. I fixed this problem by mainly created an object with very few polys and knifing only the surfaces that I needed to change. This way no extra polys were added to my models.

John:

Lighting and shadows were major issues with this scene.  What with Chris, Alisha, and Ross going nuts with modeler (the "Mod Squad", I call 'em) effectively driving the poly count to over 500,000, render times were patently un-acceptable using the old "crank all the settings to high" approach.  A lot of cheating was called for in order to bring the render times down to something approaching reasonable.

The first trick is to minimize shadows.  With eight lights in the scene, six of them from lanterns, the amount of shadows thrown, and the extent of them, produced a crazy-quilt of splotch patterns on the walls, the floors, everything... so the solution was to make every light source two independent lights right on top of each other.  One light provides the luminence and the other, much weaker, casts shadows.  The combined effect is a bright light that doesn't throw strong, expensive-to-ray-trace shadows around with all the subtlety of a professional wrestler.

Another skip is to turn off ray-traced reflections entirely and provide the reflective surfaces with a reflection map of the scene.  Saves oodles of time.

And, of course, the volumetrics... arrrgh... eight volumetric lights in one scene (with 1/2 million polys, remember) adds up to watching your hair turn gray faster than your frames render.  Two-word solution: sprite mode.  Well, okay, sprite mode with textures applied, but you get the idea.  Viper is your bestest buddy in the whole world when setting up volumetrics of any kind, too.  Add some lens flaring to the lume-light and no one will know that they're not 'real' volumetrics!  Hah!

Unfortunately, due to equipment hassles (read: my computers kept crashing) I wasn't able to put other tricks to the test, such as baking the radiosity (grr) or attaching glow shaders to the lanterns, or Fresnel shaders to the shoji screens... oh well, maybe next time.

Take-home point:  There is NO reason, ever, to trust Lightwave to provide high-quality renders simply by cranking quality settings.  These scenes were produced with a fair amount of 'cheating' and they look fine. </me toots the group horn/>

Ross:

The katana was constructed by first making a cross-section of the blade with the pen tool, and then using a rail extrude along a curve of three points. The knife tool was used to generate more polygons for the point of the blade, and points were moved one by one. The handle and pommel were simple discs with the bevel tool used to add more detail.

The teapot was made by placing points along a background reference image, making an open curve and then using the lathe tool. I used too many points in retrospect, generating wasteful superfluous polygons and making the spout harder to make. Making the spout was a difficult and clumsy learning process involving moving, rotating and stretching many points and polygons. I used subdivide and weld when necessary. When I finally got the main spout created out of the pot, I used the bevel tool to make it hollow with a lip. The lid is a disk repeatedly beveled, and the handle was made using a disk and rail extrude.

The shoji screens and moon window grid were straightforward to construct using boxes and repeating them with the array tool, although I had to be careful with the measurements to make things come out even.

Textures for the walls and screens were generated with photoshop, and the tatami mat texture was modified in photoshop.

I like how the scene conveys a sense of mystery and intrigue.