PDA

View Full Version : Complex Camera Mapping Techniques?


David Wyand
05-01-2003, 03:22 AM
Greetings! :wavey:

I've been following the most recent VFX Challenge: Outpost with some interest. I've noticed that a number of participants have made use of 'camera mapping' (or fixed front projection mapping) to create some neat effects. They've been able to block out canyons and stone arches to great affect.

I've gone through the tutorial at the NewTek site, and understand the principles behind it. I've even read through the appropriate section in the LW manual and Dan Ablan's LW 7 book. Each of these examples (and any others I've found on the web) is too basic to accomplish what has been done by others in the Outpost VFX Challenge. They all relate to placing simple objects, such as boxes or planes, to block out the scene. Or, knowing the exact dimensions of the objects in the projected picture. None of this is the case with the images from the Challenge.

The problem is that once I've setup the perfect camera position, rotation, and zoom to match the projected picture, how do I then go about modeling the object to project onto, such as a stone arch? I've yet to figure out how to transfer my camera into Modeler so that I may model according to its perspective. And that appears to be the stumbling block for me.

My question to the LW community is: How does one accomplish complex camera mapping in LightWave? :shrug:

Thanks in advance.

- Dave
http://www.gnometech.com

Steve Warner
05-01-2003, 03:36 PM
My understanding is that camera mapping is achieved in Layout, not Modeler. The technique works best on simple objects, such as desks, walls, etc. Things for which you can import a box and stretch, move, and rotate it into position. From there, you apply your Front Projection and then save the transformed object and/or bake the texture. Worley has a plugin that facilitates this in their Taft collection. I've got it, but to be honest, it's far from intuitive. A better solution is one of the camera matching programs out there. ICARUS is widely used in the challenges, primarily because it was free. It features both camera matching and photo modeling applications, and works well with Lightwave. Unfortunately, it is now being developed into a commercial product and is no longer available.

If you can't afford Taft and can't find a copy of ICARUS, this tutorial might help out:

http://www.geocities.com/kevman.geo/l65_tut1.html

Cheers!

Steve

David Wyand
05-01-2003, 04:15 PM
Greetings!

Steve, thanks for the reply. As you've pointed out, camera mapping only in Layout works for simple objects. And that I've tried and have had success at.

However, I was after how to camera map with more complex scenes, such as what a number of people have been doing for the Outpost VFX Challenge here on CGTalk. Here's an example of what I'm looking to recreate (taken from Equinoxx's VFX Challenge (http://www.cgtalk.com/showthread.php?s=&threadid=50013&perpage=15&pagenumber=1) ):

http://equinoxx.cgcommunity.com/challenges/vfx_outpost/wire_overlay.jpg

He's managed to model the various hillsides; a non-trivial task to be sure. Unfortunately, this was done in another 3D package.

To perform the same task in LightWave, I feel you'd need to move your mapping into Modeler, as you need to create new geometry based on the image itself (and the camera angle you've figured out from the picture).

I've taken a look at Taft (website only; I don't own it), and it appears it would help out with the initial calculation of the camera angle for the mapping, but I don't believe it would help out with the building the actual geometry. As with all LW camera mapping examples I've come across to date, it requires that you already have the geometry.

ICARUS has come up in the past, and as you've said, it has gone commercial and is no longer available. :thumbsdow I'll see if I can find an older version an see if it does what I'm looking for.

In the mean time, I'd really like to do this all in LightWave, if at all possible. Unfortunately, all of my attempts to date haven't worked out. :banghead:

Thanks for your help.

- Dave
http://www.gnometech.com

uncommongrafx
05-01-2003, 04:43 PM
Dave,
Describe what you've done so far.
My thoughts are that all you have to do is pull that map into modeler, decide on a scale, and put polys over what you want. From there, back to Layout to front project. Obviously, you'll have to guesstimate on the depth of things but that seems it would be trivial.
If you look at the polys it's all very low detail.

There's even a texture landscape maker that you could map that image onto and this plug would make the terrain for you.
Robert Wilson
UnCommon Grafx

policarpo
05-01-2003, 05:01 PM
check out these samples here.

they are very illuminating.

even though they are done with Electric Image, the concept will translate to LightWave.

http://www.dvgarage.com/market/product/product.php?prod=maplab&sub=samples

Steve Warner
05-01-2003, 05:06 PM
I'll be real honest and say, "I don't know" on this one. But here's my guess:

It looks like the backplate was composed from different photos. If you follow the thread, you can see that the backplate changes as the challenge progresses. I'm guessing that he roughed in some geometry using each of the original photos. He probably did this using the images as background templates in his modeling package. Once the final backplate composite was completed in Photoshop, he probably took those rough models and tweaked them to match. It looks like the models were placed in the scene as separate objects. This helps give the feeling of parallax when the camera moves. From there it looks like a "simple" matter of matching the perspective.

Equinoxx said himself:

it's so easy that eventho it is my 1st time ever doing this, i did not need a tutorial.
- make sure the you match the camera you're going to use for cameramapping as close as possible to the perspective of the image.
- Model geometry stand-ins as accurate as possible for use with camera mapping.
- don't overdo the movement in the camera you are going to animate.


Now that you've got my interest piqued, I'm off to try this myself. I'll let you know if I have any success....

Cheers!

Steve

David Wyand
05-01-2003, 05:21 PM
Greetings!

What I've tried to date is to first match Layout's camera's position, rotation, and zoom to the background plate. In the picture above, it appears that the camera is pitched downwards, and with perhaps a slight wide-angle (I haven't actually tried to match this particular picture, and only put it up as a completed example).

The next step would be to then take this camera into Modeler, decide on a scale, and begin to build geometry based on the camera's view. This is the step that I've gotten stuck on as it's not possible to transfer the camera information from Layout into Modeler. I've tried to build a stand-in plane that I've mapped the background to based on the camera angle and zoom that I then bring into Modeler as a reference, but so far that hasn't worked out either.

Robert, I hadn't tried your suggestion of just starting in Modeler and not worrying about the camera setup. My gut told me that there'd be a number of problems with this later on when it comes time to have 3D objects interact with this environment. With the terrain built from an orthographic view, and the 3D objects being shot from a perspective view to match the background, I'd have thought that it would be difficult to line everything up (for shadows, reflections, etc.). I'll give it a try and let you know how it works out.

Have you actually tried this out yourself, and if so, could you please post an example?

And please elaborate on the terrain landscape maker. I'd be interested to see how a program would automatically extract the vertex information from an image to build a terrain.

Thanks.

- Dave
http://www.gnometech.com

HowardM
05-01-2003, 05:27 PM
heres a TIP!...you can model in modeller, but see your results in layout!
using the HUB, and a 2 monitor system, you setup your image in the background in Layout, and then you start with a single poly in Modeller...as you tweak the polys, Sync the object and youll get an update in Layout...start adding polys , moving and twisting in 3D until youve what you want...I know tedious, but it works!
:)

David Wyand
05-01-2003, 05:30 PM
Greetings!

The above post was meant for Robert. Seems you two guys posted the same time I was writing a response to him.:p

policarpo: Thanks for the link. I'll take a look at it this afternoon.

Steve: Yup, using different photos for each layer is definitely something you can do. It can certainly make for some interesting composites. From the quote you posted by Equinoxx, I understand the steps he's done are as follows:

1. Match the camera to the photo.
2. Model the geometry to be mapped.
3. Animate appropriately.

It's the transition from step 1 to 2 that is tripping me up as I cannot build my geometry from the camera's perspective. Hopefully policarpo's link will help out, or Robert's suggestion will work.

Thanks.

- Dave
http://www.gnometech.com

froggyplat
05-01-2003, 05:51 PM
i'm interested in trying out some of these techniques too. thanks for the info.

btw: here is a link i found with a lot of hi-rez photos of the Grand Canyon, similar to the shots being used in the Outpost challenge.

http://nathancheng.com/grandcanyon/

uncommongrafx
05-01-2003, 06:25 PM
I'm willing to do it on something you have and would like to see done.
I've done it mostly based on tutes in learning the NEW lw but I've done a fair share of terrains with that plugin mentioned.
Actually David, the problem that you envision with it not matching is exactly why the projection will work!

Someone somewhere made the comment that you can project with one camera and view through another...

Texturescape is the name of the plugin.

Froggyplat, thanks for the link!! Gonna sit and dwnld them all...hehe.
Robert Wilson
UnCommon Grafx

Cman
05-01-2003, 06:27 PM
Awesome link on the grand canyon shots!

I'm looking for a link on a tut I just watched yesterday that covered this technique.

edit
I believe this is the tut I saw. My server is really slow today so I can't get this thing to download. :annoyed:

2-1/2 D (http://www.floridafx.com/Training/Tutorials/25DFX/Why_2-5D/Why_2-5D.html)

If it's the one I remember they use LW to build a 2.5D comp.

MorBioS
05-01-2003, 06:42 PM
great great great!!!!!


more more more tut links:bounce:


very cool links policarpo and cman:thumbsup:

mbaldwin
05-01-2003, 07:58 PM
Newtek: please, for the love of God*, solve camera mapping for lightwave!

the split between modeler and layout sucketh the most with camera mapping. The split used to really blow with rigging characters, but now there's lots of 3rd party tools to ease the pain.


*drama denotes urgency of request.

Kvaalen
05-01-2003, 08:31 PM
That link with the canyon shots rocks!


Newtek: please, for the love of God*, solve camera mapping for lightwave!

the split between modeler and layout sucketh the most with camera mapping. The split used to really blow with rigging characters, but now there's lots of 3rd party tools to ease the pain.


*drama denotes urgency of request.


I hope LightWave doesn't get put into one package, I love it when it is split.

HowardM
05-01-2003, 08:39 PM
...so wasnt there a plugin or a method used to model in Layout where nulls or bones are assigned to points, etc, etc....or cant this be done by hand?
hmmm...make a grid of about 100x100...assign bones to all the points...then push and pull your polys into the right shape in Camera view......then Subd?

hmmm?
:)
...but really, wasnt there a plugin?

mbaldwin
05-01-2003, 09:42 PM
Kvaaleen,

i don't care if it's one app, two, or twenty. Just fix the workflow. With this topic, bring cameras into modeler(cameragons?--sheesh), or modeling tools into layout, or something.

Lightwave is the most accessible app for the lone-gun operator. fixing camera mapping would open up a whole new industry to Newtek: that of illustration, photography, retouching. Especially when the 3 professions are asked to use more and more stock art in their final sollution(art where they're not privy to the camera/location conditions easily enter numerically).

Peter Jackson showed how with Maya, he was able to extend camera moves on The Two Towers, going higher, lower, or farther than what was originally filmed. What still artist wouldn't jump at having that same capability with their own work? I know I sure could have used it on many jobs.

policarpo
05-01-2003, 09:45 PM
Originally posted by mbaldwin
Kvaaleen,

i don't care if it's one app, two, or twenty. Just fix the workflow. With this topic, bring cameras into modeler(cameragons?--sheesh), or modeling tools into layout, or something.

Lightwave is the most accessible app for the lone-gun operator. fixing camera mapping would open up a whole new industry to Newtek: that of illustration, photography, retouching. Especially when the 3 professions are asked to use more and more stock art in their final sollution(art where they're not privy to the camera/location conditions easily enter numerically).

Peter Jackson showed how with Maya, he was able to extend camera moves on The Two Towers, going higher, lower, or farther than what was originally filmed. What still artist wouldn't jump at having that same capability with their own work? I know I sure could have used it on many jobs.


i haven't tried it, but wouldn't this work:
http://www.worley.com/taft/taft_cameramatch.html#topcameramatch

mbaldwin
05-01-2003, 10:06 PM
Policarpo,

I have it. It's better than nothing, but doesn't approach the ease of dragging points in space. Also, we probably need some sollution that works for more organic things, like landforms. Being able to fashion rough cages directly over your camera view would be helpful.

All of this is do-able in lightwave, but hardly efficient. As an illustrator, there's a fine line between making or losing money on a job. also making sure that most time is spent on the creative process, not the mindless mechanical necessities. One could also argue that numerics don't live on the same side of your brain as the creative stuff--minimizing the math probably saves commuting time/stress between lobes.

policarpo
05-01-2003, 10:09 PM
Yeah...i can understand that.

I usually composite in Photoshop and use the power of Photoshop to do my tweaking...but yeah...having a powerful Camera mapping tool in Layout would rock!

I'm still new to camera mapping in LightWave..so I don't know how easy or painful it is to setup. :)

David Wyand
05-02-2003, 08:37 PM
Greetings! :wavey:

First of all, I wanted to thank everyone for replying to this thread with their suggestions. I've looked at every one and have thought about what they (and you) have said.

Yesterday, I decided to try and build out a stone arch based on the procedure Robert mentioned above. I'll first describe what I did, and then go into some issues I've had. There's even a movie in it for ya if you read all the way through! :)

I first began with finding a suitable picture from which to model. I wanted something fairly complex so you couldn't just build a couple of boxes and be finished. A Google search produced a nice stone arch (which you may download from my site here (http://www.gnometech.com/lw/cgtalk/cameramapping/stonearch.jpg) if you wish to also give it a try).

I then placed the image into the backdrop of the Back view in Modeler, and began to draw polygons. I mentally divided the arch into four sections: back of arch which flows into the top; side of arch; underside of the arch; and the ground. It was then a matter of deciding on the contours of the polygons and how much detail I wanted. I had also decided that I would build the object in two layers: one for the stone arch; the other for the foreground.

http://www.gnometech.com/lw/cgtalk/cameramapping/archmodel.jpg

During the modeling, I also had the object loaded into Layout. I had a camera set back a few thousand kilometers and zoomed into the object to try and flatten it out against the stone arch backdrop. This camera was used for mapping the image onto the arch object. I set it up this way to try and mimic the orthographic view that one has in Modeler.

After a lot of vertex pushing, I came to the following:

http://www.gnometech.com/lw/cgtalk/cameramapping/progress.jpg

The left image is the original stone arch picture. This middle image is what the arch object looks like through the camera that was used for mapping. The final image is rendered from another camera's perspective with the stone arch picture projected onto the LW object I created. It also has a LW generated gradient sky.

The final step I did was to bake the colour map onto the object so I could move it around (specifically the foreground layer, which need to be moved up for my shot). It was then time to animate the camera and produce a movie:

QT 6.1 Movie of Stone Arch (3MB) (http://www.gnometech.com/lw/cgtalk/cameramapping/stonearchsound.mov)

So, in the end, it seems to have worked. However, there certainly were some pitfalls with this method. I've already written enough here, so I'll write about the problems in another posting.

- Dave
http://www.gnometech.com

froggyplat
05-02-2003, 08:49 PM
wow, looks really cool!

any chance you might post your scene files so we can take a peek?

HowardM
05-02-2003, 08:55 PM
Sweet!
So did you try modelling in layout like I suggested using the hub?...I dont see how you can do it in modeller alone...
How long did it take to push and pull those verts?

I wonder how many effects use CM techniques....its something we all should use more often!!
:)

Cman
05-02-2003, 09:09 PM
I wonder if you could save time using a separate camera parented to the object, just to hold the texture, instead of baking.
Would the extra camera slow your workflow, do you think?

comanche
05-02-2003, 09:16 PM
Hi Dave,

I'm wondering why you set up the camera in Layout the way you did (extreme zoom etc.). Wouldn't a flat projection do the job in this case? You modelled in back view, so flat projection is all you need IMO. Or am I completly wrong?

cheers,
comanche

David Wyand
05-02-2003, 09:49 PM
Greetings!

froggyplat: I'm glad you like it. I'll post up my scene and objects shortly.

HowardM: I did sort-of use your technique. Rather than using two monitors (which I don't have), I did a constant ALT-TAB between Modeler and Layout to compare.

Cman: I believe you're right. I could have one projection camera parented to the stone arch object, and another one parented to the ground object. That way when I moved the ground object up for the animation, the texture would have moved with it. Good idea.

And finally, the big response:

comanche: Uh, DOH!:argh: You are absolutely correct. Based on how I modeled the stone arch, I could have just used projection mapping down the Z-axis within modeler and avoided camera mapping all together. I was so focused on trying out camera mapping on a complex object, I completely lost sight of the simpler solution. Doing everything in Modeler would then eliminate the problems I was just about to write about (i.e.: The camera mapping not lining up with the model as no matter how far I pulled back the camera and zoomed, I could not exactly match the orthographic view...thus fringing would occur).

Now, all that said, I am working on something to demonstrate when you would need to map through the camera's viewpoint. At least that's the only solution I've come up with so far for this case (and perhaps someone else will have a simpler solution :) ) I'll hopefully post it later today.

- Dave
http://www.gnometech.com

comanche
05-02-2003, 10:07 PM
Flat mapping might work with only a few images (like the stone arch you've just used, Dave). So I'm really looking forward to the solution you'll come up with. Thanks for sharing all this, Dave.

cheers,
comanche

David Wyand
05-03-2003, 04:05 AM
Greetings!:wavey:

I've just archived my scene and object files of my stone arch for all of you to use as you wish:

Stone Arch LW Scene and Objects (http://www.gnometech.com/lw/cgtalk/cameramapping/cameraprojectiontest.zip)

As was pointed out above, there is probably a better method to arrive at the final mapping solution for this instance than the one I presented. However, as I started this thread to learn how other people accomplish this task in LW, it's all good :thumbsup:

Below is a description of what's in the archive and how to use it. My next posting will be in the next day or two to show you how not modeling through the camera's lens can get you into trouble (i.e.: problems with Modeler-only solutions).

Archive Contents

ArchSetup02.lws - This scene is what I used to set up the fixed projection mapping (or camera mapping) for the stone arch I built in Modeler. When building the arch point-by-point, I would flip back to Layout to make sure that everything lined up with the projection. In Modeler, I used the stone arch picture as a backdrop in one of the viewports as reference (see my Modeler screen shot posted above).

The other cameras in the scene were used only to move around the model to check it out, and were not used for the final rendering. Once the model was complete, I set up the Surface Baker for both the Arch and Gound layers (which I separated to allow me to later composite the shot).

ArchBaked.lws - This scene makes use of the two surface baked layers for the final animation. The camera movements were designed to minimize distortions due to limits of projection mapping, as well as to hide any flaws in the modeling.

ArchAnimMaker.lws - This scene is used to assemble all of the rendered images from the above scene into the final animation. I prefer to render to images rather than straight to the final anim as it allows for more control.

For example, I found that the Sorenson Video 3 codec for QuickTime really darkened down the images. By using the HDR Exposure Image Filter with the black point dropped down, I managed to lighten the whole anim without needing a re-render. LW can be used for quite a few compositing tricks for those that don't have dedicated compositing software. I'm sure this isn't new to most, but I thought I'd share my workflow for those that may be curious.

Enjoy!

- Dave
http://www.gnometech.com

comanche
05-03-2003, 10:56 AM
Thanks a lot for the scene files and the explaination, Dave! :thumbsup:

comanche

froggyplat
05-03-2003, 01:40 PM
hey, david, thanks for sharing the scene files!

aurora
05-03-2003, 01:58 PM
Man, this is one of the best threads I have read in a long time. I have often wondered how they accomplish this two but never made the attempts to try it. David thanks for the work and more importantly the mini-tut and files. I'm downloading them now and will give it a try this weekend. If this works man have I pic I have been dying to try this on. I'll let you guys know how it goes.:applause:

HowardM
05-13-2003, 02:30 PM
This new MiniMo plugin looks like a big help for this technique!
http://www.infoseek.livedoor.com/~f_ichikawa/

David Wyand
05-14-2003, 05:20 AM
Greetings! :wavey:

I'm sorry that I haven't yet responded to this thread as I promised. I was hoping that I'd get some things done before I went on my trip, but alas time wasn't kind. Hopefully I'll be able to continue this thread when I return home next week.

HowardM: I'll be sure to try out that plug-in when I get back to my computer. It looks interesting...

- Dave
http://www.gnometech.com

CGTalk Moderation
01-15-2006, 12:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.