I had noticed this as well, when I tired to make my own trees from images vs. formfont 2d trees.

At the time I didn't have good solution, but since then I've used LIveTrace in Adobe Illustrator CS3. That could be good solution. You could load hand sketch (or image file there) and have it create the outline which you can export as dwg/dxf file. Then take your tree texture in SU and paint it on your newly made surface.

I wish it could shadow shapes in transparent PNG,s but don't think that it can do this?

I don't care so much what the edges are like. The point I was making, is that my goal of my process was to be able to find the center of the photo without just measuring to the middle from the sides. I find the center purely on geometric markers. This allows me to see if a photo was cropped and which sides.

The less re-sampling of your source photo, the better. SketchUp is already going to have to re-sample the photo (when applied to surfaces), no reason to have Photoshop do it as well.

For photos where your aren't looking up or down (parallel verticals) I do need to do a little more research. While I thought I resolved it and modeled geometry looks OK, I have my thoughts that maybe not accounting for side cropping correctly could affect measurements on the building.

By the way, I also believe it should be pretty easy to model one point perspective. I'll post on that another time.

Well, making things vertical isn't really enough. See Picture 6 attached. If you also play with the horizontal perspective to get the horizon level at the same time and the verticals, and then you adjust the canvas size to get the horizon in the center, then you can do it this way; see Picture 7 for screenshot after additional adjustments. But that is also a lot of work, and requires distortion of source image. The method I used just expands the canvas, and doesn't degrade the image.

My test image on previous post was not cropped. It was a test image where I know the center, but the method I used to find it doesn't rely at all on the edges of the photo. If you do same analysis on small piece of the image, you will find the same center, may even be off the edge of the photo.

OK, so I expected to find a lot of answers on the internet for how to find the center of projection (or attention) for a photo, but surprisingly I didn’t find any techniques. Ok there are techniques that look at peoples eyes to see if they have been composed into a photo; this work through finding the center base on analysis of the eyes - the bad eyes will project a different center than the rest; but not really relevant for PhotoMatch.

Well two things came to mind that I thought would be useful for this. First is that there isn’t just one horizon line, but rather three (for cube shaped object – 3 90deg axis). Second is that each horizon although it goes off infinitely in both directions, kind of has a center. This center is the point along that horizon, where a vertical line doesn’t tilt in or out regardless of tilt of camera. It is perpendicular to the horizon and goes through the third vanishing point (the one not forming this horizon). Ok, a third thing was that as you tilt your camera up, your verticals start to converge at the top, and your horizon moves lower on your photo. the question was how to translate the amount of vertical vanishing to a camera tilt amount; this would seem to vary based also on the FOV of the lens.

So I drew this line for the three horizons, and wouldn’t you know it, they all cross at (or very close to) the center of the original photo. You can crop the photo all you want, but this point won’t change. Of course you need to have content in the photo to define the vanishing points.

The attached photo (pic 3) shows me drawing the lines perpendicular to the horizons through the opposing vanishing point. This particular test was off about 3% from the true center, but keep in mind, this error may be the camera’s sensor, they are always off to some extent, although 1% would be more typical. The other flaw is that I may not have drawn my lines correctly do to use error, or other lens distortion. But I think its close enough to provide good results. Had this photo been cropped off center, I would now open in Photoshop and expand the canvas with extra white space on the sides making sure to re-center on the newly found center of projection.

So it seems this works great when you have the camera tilted up or down (where there is a vertical vanishing point), but when you are truly or close to being level, your vertical vanishing point is so far off that verticals appear parallel to each other. Well, that doesn’t let me get my center of projection. I know it is on the x-y horizon so I can resolve vertical cropping, but I won’t be able to correct for cropping to the sides of the photo in this case. This doesn't seem to be an issue for PhotoMatch though. If you don't have converging verticals, then you don't have to worry about cropping to the sides. It is probably a good idea however in this case to rotate the photo so the horizon is level. This way no-matter where the center was across the horizon, it is still vertically in the center of photo.

Happy un-cropping. I'm hoping to not find the need for this in SU7, and also hopping for mostly automated calibration as well. Maybe also option for telemetry based solution to work with non-square and non-adjacent geometry. (but I'm just guessing and wishing)

Feel free to post problem images, and let me or others take a stab at finding the center.

Didn't see any software mentioned in other thread, but here are two alternatives, but I know there are a number of others.

Realviz Image Modeler (now owned by AutoDesk). Very powerful, hard to learn, expensive.
Canoma (Metacreations sold to Adobe in late 90's, if you are really lucky, can find copy on ebay) This is very easy to use, and very powerful (but not nearly as good as Image Modeler for larger scale projects.

Both of these use pins on multiple photos so can deal with more complex geometry. Makes no assumption about anything being 90deges. For these or any software, lens correction is critical.

I'm sure there are others, but don't know much about them. Here is one.
http://www.photomodeler.com/index.htm

Another category of software, more for shooting small objects; you place object on target with dot patterns, and shoot from multiple angles. Modeling is then mostly automated.

You can stack scenes with different geometry hidden in Layout, but that brings up its own problems.

If you are really only modeling in 2d you could move objects that you want a lighter line weight very far away and use depth cue setting.

And yes, exporting to Illustrator also works. With CS3 you even have ability to export and import dwg/dxf to scale.

Jorensen. Can you post picture or describe how it looks terrible. Could be related to image format, resolution?

Caronte01; as far as the fading. I believe I've mentioned this to SU, and was just the way watermarks were setup between scenes. At least when you change watermark from one scene to the other. I'm not sure if it does this if you have same watermark or style across different scenes. I could test that, if that is the issue.

Ideally I'd like to either have a hard cut, or a cross dissolve, vs. fade out/fade inn. But in perfect world could make this user preference. (keep in mind I'm just talking about behavior of watermarks vs. the model iteslf.)

I use VMware vs. Parallels. Used to use Parallels, but VMware just seems better in a number of ways (stability, speed, features).

But I didn't have to use between VMware or Bootcamp. You get both. I can launch VMware when I went it and mac open at the same time. And I can reboot into windows when I want real power. I think Parallels does the same. I will say that I have issues with startup time for Windows (others in office don't have this problem, have not pinned down why), but once windows is running, application are fast.

One things I can't do with dual vm/bootcamp setup is to suspend windows. With VMware only you can quit VM ware where it saves your windows state. Makes for quick startup next time, because windows is right where you left it with applications already open. Since mine is shared with bootcamp, it has to be shut down when I'm done.

Yep, with little surfaces here and their coplanar, vs. an entire building or building facade, I can see that being a pain.

Maybe SU 7 can add a "search for coplanar surfaces" option. I would thing some modeling apps would have this, but maybe easier with solid vs. surface modeling. Come to think of it, when you import Autocad to SU, is there a checkbox for "merge coplanar faces". I guess you are using that, but just not working all the time. (I can't access older threads in this post while I'm writing"

I'm on a mac in an office that is mostly PC (maybe 60pc, 5 macs?)

I'm using Kerkythea. Is free but can be very powerful. Outside of quality you can get from it, big thing for me outside of its integration with SketchUp was that it is both mac and PC. So if you have big job to render you can dump to PC while you keep working on the mac. It will do network rendering, but I haven't used it for that yet.

I would not recommend rendering under parallels or under vmware. If you are going to render in PC. Either use a different computer, or boot to windows using BootCamp. You don't want to render with you computer resources split running two OS's.

Lot of other rendering apps for mac, so KT is by no means only good option, but just worked well for me, and i liked the active forum and development of the product.

Thanks, I think I thought of that from 5th element or some other movie where they were building up a person, bone first, then tissue, then skin. The trick was doing it with just two scenes vs. something more complex. That is when I realized I can adjust the heights of the lower and upper section cut for each group as a timeline to control the start and finish of the build. As long as what you want is linear (e.g. no speed ramp up or down) you can get any timing you want with just two scenes.

@plot-paris said:

very impressive, David.

it never occured to me to use several section planes to show different elements of one building grow.

thanks for the demo file

If you have 3d components for these objects here is a solution.

For each object component, have duplicate versions on different layers (top, front, left, . . . , 3d). By default use the 3d view, and have other layers set to off. But when you are in top view for example, disable the 3d layer, and turn on the top layer. I've thought about this for working with trees. 3d for perspective, but plan version for plans or as working object to speed things up.

If you want an easy way to see all of these views in Layout, Just setup a paired SketchUp and Layout file with views already setup. Then just drop component into your SketchUp file, goto Layout and Adjust the scale as needed for that specific component.

Jackson, since the rendering is where you see the problem with coplanar components; seems you can use that to your advantage in trying to identify which components have a coplanar copy. Just create one or a few test renderers to identify your problem areas. Then you only have to mess with the few hundred vs. thousands.

If you are worried about trying to match up the render to the SketchUp file, just overly as a transparent film on top of the sketchUp model (maybe watermark (although that can be glitchy) or you can use 3rd party app that makes windows transparent.

If Google did by rendering engine, I expect it wold remain a separate application like Layout, vs. becoming part of SketchUp.

I use KT, and like that my rendering goes of to separate application. I can let it do test renders while I continue to model in SketchUp, and for big renderings and animations, I dump it to another computer, and again, I can keep working in SketchUp.

Now that basecamp has wrapped up, has anyone come across a wrap up blog/post. The blog is empty. Don't know if possible future features were not shown, or were shown but fell under NDA.

I used the two section technique for a campus masterplan (about 15 phases) some phases demolish a building, while others add new buildings. Same technique, just reverse order of scenes with section active.

Here are some tips

• to have consistent speed of buildings rising from the ground place top section cut for all buildings just above the height of the tallest one.
• if you want the building to look like something other than an empty box, at a minimum add in some floor plates, and double wall for building shell. Columns, building core, windows, would look great.
• you can create your scenes to setup your sections cuts, and then alter scenes for camera path (update with only update camera location set) or do camera path, and then section, but I liked to get sections set first, so I could preview final product better. I used flightpath to help create my path.
• my original thought for camera movement was to just do a circle around the campus looking towards the middle and down around 45deg. Ended up having nice look where camera moved around to approach where building was going to be demolished, and when one was going to be built. The camera started to focus on camera before demo/build started, like it knew what was coming next (which of course it did).
• In my case I created every building as a component vs. group (its my view that SU7 should just drop groups and only offer components). I initially just did a simple box for buildings, and saved out to folder of saved components. Then while I was working on the flight path and look of the project, someone else was modeling the buildings for me. Whenever they updated a building, I just reloaded the component.
• I also added a watermark for each phase that ghosted in white text throughout the animation.
• I dropped project into Layout to make 11x17 handouts with plans of each phase.
• For simplicity I color coded new vs old buildings.

Sorry can't show the project - will be secret for a while. If I get time someone I'll make a dummy version of it with generic buildings and blocks (to protect the innocent).

[edit] I added attachment of test file where I made the skin, floors, and columns come up at different times to make it look more like it was being built. I did this by staggering the heights of the section cuts for each group.

multiple section cuts to animate building going up

I haven't tested this, but in theory if your have 5 scenes and each is 2 seconds long. And you have a frame rate of your export of 2 seconds a frame (not 2fps, but 0.5fps). It could work, but could be kind of glitchy.

I'd recommend using Layout for this. Another variable this addresses is if you change the aspect ratio of your screen in SketchUp. As Layout is more of page layout application you can resize your window without effecting the output.

300dpi is highest res export, so if you need 4096pixels make sure page is wide enough. 11x17 should do it. Oh, Layout also gives you more line weight control that is independent of export resolution.

Are you on mac or PC?

On mac under preferences/startup I have three options

• Prompt for Template
• Blank Template
• Use Selected Template

Maybe you have third option checked instead of first?

Maybe have a layer lock act in one way (as it does now), but then also allow individual objects to be locked (where double click doesn't allow text edit)

You may also want to make a group out of the property box. This way you can move things around that you place on it, without affecting the property itself.