PDA

View Full Version : How to model for camera mapping without the bloody layout cam??



Taro Yoshimoto
03-26-2012, 07:57 PM
Hello everyone!

I need to model rooms over photographies to map those photos on the models. Basic camera mapping. But how the hell I am suppose to match the cam if modeler dont have a cam..? That is ridiculous.

Any ideas?

ps; how do we change the perspective view lenght in modeler again? I did it before but cant remember.

thanks

wrench
03-27-2012, 03:19 AM
Hit d and on the Layout tab you have Perspective amount.

B

vector
03-27-2012, 03:41 AM
Maybe using camera projection and using some modeling tools in Layout. In my blog I posted time ago about camera mapping and more recently about modeling tool in layout (from Japan, curiously)

http://lightwavebox.blogspot.com.es/2011/04/lightwave-camera-mapping.html

http://lightwavebox.blogspot.com.es/2012/03/3dcel-layoutmeshedittool.html#uds-search-results

Taro Yoshimoto
03-27-2012, 04:11 AM
Thanks Wrench. Sad there's no mm information. Not even a default value that you can revert after moving it. Better than nothing though.

Vector, the 3DCEL - LAYOUTMESHEDITTOOL look very cool! It's not exactly a solution but it could help with retouch in side Layout. I wonder if its animable.

What we really need would be a kind of hub between Layout and Modeler that would sync the perspective window of modeler with the camera in Layout.

MUCUS
03-27-2012, 04:12 AM
Well, as long as Lightwave will be a separate app, it will be a tricky operation to do...

But you could model the floor, send it to Layout, match the photo with Layout camera and then continue modeling in modeler (Yes you'll have to switch between modeler and Layout (F12 is your friend))

OR, you can also isolated part of photography in photoshop to make simple texture, and then map them on your model?

Taro Yoshimoto
03-27-2012, 06:41 AM
Mucus, you gave me an idea with your floor...

What about bringing few rectangles in Layout, place them all around the place on the photo to serve as guideline inside Modeler (with save transformed object) and then use this to match the perpective inside the modeler perspective view (with the photo in background)... I think it could work. Of course its an annoying process...

vector
03-27-2012, 06:55 AM
What we really need would be a kind of hub between Layout and Modeler that would sync the perspective window of modeler with the camera in Layout.

In instance: A new mode for the viewer in Modeler called "Layout camera" emulating what the camera views

stiff paper
03-27-2012, 01:00 PM
This tool might help:

LayoutMeshEdit, by 3dcel:

http://www.lwplugindb.com/Plugin.aspx?id=b2b289fe

ivanze
03-27-2012, 01:41 PM
Watching erikals video of Layoutmeshedit makes you think that the hooks for modeling tools in layout are already there.

Mr Rid
03-27-2012, 05:57 PM
Since I do a lot of CG integration with plates, this has been the biggest problem with having a separate modeler. I often need to create geometry (rather complex at times) to match a plate, and have to once again resort to ridiculous workarounds for something that should be basic. If I can import a layout's camera view into a third party compositing app, I dont see why it should be difficult for NT to address this in Modeler.

JonW
03-27-2012, 08:20 PM
Put photo in backdrop in Modeler (4096 x 4096 is the max size & the best size to work with if you need all the detail).
Make an object, outline key parts of the image which need to be aligned in a 3d object (if you make the object a known size, what I do is for 5616 x 3744 pixel image is make the over all size of the object 561.6 x 374.4 mm & also put some corner alignment marks in the object).

Layout:
Parent that object ("m" Motion Options) to the camera. Scale the object down & move it in very close to the camera on the Z axis (you can calculate this to the pixel if needed). (See attached image with the camera angle made a touch wider that the blue outline). (If you don't need to render the outlines but keep the outline in place in Object Properties/ Render/ Tick all the "Unseen" & Un-Tick all the "Shadow" boxes)

Then you can move & rotate the camera to align with your 3d object.

Then you can project the image & it will align with the 3d object.

For accurate alignment it's best to correct camera distortion in the photo before you get going.

Attached: the blue outline was traced from the site photo in Modeler. In Layout the 3d objects were massaged to fit the blue outlines. There are some differences here because the survey information didn't match up.

Taro Yoshimoto
03-27-2012, 11:36 PM
I'm afraid to say, JonW, that I am quite confused by what you wrote. Could you please tell us the general idea before going into details?

JonW
03-28-2012, 01:39 AM
If you have a photo of the room, trace the corners of the room etc, & make lines or outlines (the Blue Lines in my image above are each individual objects made into one large object in Modeler but with no depth (Flat plane). The outlines are on a flat plane, & that flat plane object is placed in front of the camera, I scale down the object to 20 or 10% etc & place it a few 100mm in front of the camera's (depending of focal length) Z axis after it is Parented to the camera so you are looking through the Blue line object. (some people place a bit of a Acetate film on their monitor with outlines drawn in Texta from their photo)

Build the room in 3d so it's a guess of what dimensions it should be. In Layout your camera will have the Outline object in front of it, Parented, on the Z axis a few 100mm or so. A bit like a "Gobo" but it's in front of the camera & not a light.

Then move & rotate the camera into position so the 3d object is looking sort of in the right position. Then adjust the size of the 3d room to fit the outlines. You may have to go back & forth a bit moving the camera a touch & then resizing the room. If you know the position of where the photo was shot from then you can move the camera to that position & get it pretty right first go. (& knowing the focal length of the lens helps as well) You will need to adjust this as well as position & rotation.

It's exactly the same as lining up a scene for a 3d photo montage, just that you are doing the inside of the building & not the outside.

If you are still having problems post the picture you want to use & I can set up the basics for you.

Taro Yoshimoto
03-28-2012, 02:22 AM
Ok, I think I follow you now, but Im not sure you are follow me. The method for matching the object to the photo was not really the point of this post. The question is more how to deal with the lack of real camera and focal lenght in modeler and avoid loosing too much time because of that. Lets say that my photo was shot in 17mm... the perspective slider in modeler prefs would be highly innacurate to match it.

As for the method; I normally composite the photo in background and model in wireframe over it. Your method is more fancy though, no doubt about it.

Another method I use is to import a bunch of geometrical shapes in Layout and work from there, with the real focal lenght and cam position/rotation. Scaling and moving my objects to fit the shape of the room or building.

If I need a lots of details for a building, I extract this part of the photo, bring it in Photoshop and with the corner distord tool, destroy the perspective to a square front view. That way I can model each windows and details in modeler and apply the square texture on the front of the model. Give me stellar results.

But yeah, I though maybe there was a tool or plugin that I had overlooked and that could have facilitate the process. Layoutmeshedit might be this very tool.

By the way, there's only 1 plugin on the 3Dpage cell... no more Layout mesh Edit...

stiff paper
03-28-2012, 02:36 AM
By the way, there's only 1 plugin on the 3Dpage cell... no more Layout mesh Edit...

It's on the main page, just scroll down to 2009-04-06.

However, reading some more detail about what you're trying to do, I think maybe it's not what you need. LayoutMeshEdit is great for pulling around points in Layout, sure, but it won't give you an accurate model at the end, it'll just give you a model that fits the BG image in that particular camera view. Great for putting a small camera move on a static image, or altering an object to fit in a particular "hole", but not that much use for very much else.

Mr Rid
03-28-2012, 04:41 AM
...

Another method I use is to import a bunch of geometrical shapes in Layout and work from there,

Thats one way- make the object out of primitives and save transform on the mess, to use as a guide in Modeler.

The real fun is modeling something organic like a landscape for things to collide with and cast shadows on or for instances to hug, especially when that landscape is only ever viewable in part as the camera pans http://www.box.com/shared/static/iktbntek1mc2l2zrzlhx.mov I may model with bones, and save xform with the hub, going back and forth.

I stumbled into one way to edit a mesh in Layout by applying ClothFX for one frame, then use the edit tools to sculpt.
103004 Allows you to fine tune a shape to line up precisely with an image- like for trees in this example - http://forums.newtek.com/showthread.php?t=104227&highlight=squirrels

Thought about doing a little tut on the particular challenge of extruding a landscape out of this plate with crushed depth http://www.box.com/shared/static/ngyvfqsm5r.mov but figured only about two people would ever need to do such a thing.

Taro Yoshimoto
03-28-2012, 05:31 AM
Thanks Mr Rid, very instructive and interesting exemples. I'll do the job with the shapes directly in Layout this time. This is going to be the fastest way for what I need.


It's on the main page, just scroll down to 2009-04-06.

Thanks Cardboard! I will check it right now.

thanks!!

ps; I notice that people are posting questions on 2 forums sections; LW community and General support. There should be a rule to put all the LW related questions in 1 place.

Eagle66
03-28-2012, 01:33 PM
Thought about doing a little tut on the particular challenge of extruding a landscape out of this plate with crushed depth http://www.box.com/shared/static/ngyvfqsm5r.mov but figured only about two people would ever need to do such a thing.

I do such things :) and asked here (http://forums.newtek.com/showthread.php?t=125266) - how to model a landscape mesh from real footage in Modeler for Camera Projection or the example above for Crowd Sim Ground.

Hints for extruding a landscape out a plate would nice...
Thanks!

Mr Rid
03-29-2012, 12:21 AM
I do such things :) and asked here (http://forums.newtek.com/showthread.php?t=125266) - how to model a landscape mesh from real footage in Modeler for Camera Projection or the example above for Crowd Sim Ground.

Hints for extruding a landscape out a plate would nice...
Thanks!


So you're the other one. :) An explanation http://forums.newtek.com/showthread.php?p=1232483#post1232483

Mr Rid
03-29-2012, 06:53 PM
Another example where I had to create a ground for objects to collide with. Again, I arranged primitives in Layout, saved transform (theres a few plugs for doing all at once), and used as ref for Modeling.

103048

The explosion was done by Rif Dagher with Thinking Particles.

Axis3d
04-02-2012, 03:15 PM
All great methods.

The method I have used in the past is to bring in a basic object or subdivided ground plane. Then run SoftFX with the Operator set to None. Basically, this captures the points in the object. Then on the EditFX tab, turn on EditTool. This allows you to push and pull the points around in Layout. Then save Transformed. It works better with a really low-poly model. Then you can add more geometry in modeler if you need.

pilF
05-24-2012, 11:17 AM
Pfmatchit gives you nulls from a tracked shot which can be used as a guide in layout to place basic geometry. There's also a null to points plugin to make modelling easier. Or you could donate some more money to the Pixelfarm and buy Pftrack, generate geometry within the track. Export that including the uv mapped surface.Only motion pictures though.