Possible facade mapping workflow

Hi
yesterday we ve been up all night testing a new workflow to get a quick facade mapping.
So you first start projecting a grid of dots over the facade, and take some nice picture from different angles.
The you may use photomodeler to elaborate the pictures and work out a gross mesh of the it. Of course there are other softwares that can do that. Even a free one, can t remembe the name right now.
Now you have a .obj but it s polluted with extra stuff so you need to import it in Maya, Blender, Max3D, etc. and clean up the whole thing to get just the polygons you need.
When you are done with that it’s time to make it vvvv friendly, we were not able to get a working file from Maya (.dae, .x) until we installed the Superinstaller from Polytrans that creates and exporter for Maya and got a perfectly workable .x file.
Import now with xFile in vvvv and be happy with a projector node.
I don t know of how much interest this is, but as we fine tune this system I ll keep posting and eventually do a mini tutorial (if anybody interested).
I also think that Polytrans should be mentioned as a possible way to export .x from Maya in the doco.
cheers

insight3d but couldnt get it to work with my camera - its still in early beta. Actually its more about generating point clouds out of pics.
Imagemodeller from autodesk which isnt shipped standalone anymore.

Havent tried it for facades, but its a quite easy way of getting realworld into 3d. Even some nice motiongraphics can be done by these tools. Thanks for sharing your workflow.

Some updates and tips.
So far I ve found out a rather stable workflow.

After getting your model with ImageModeller or Photo Modeller, just load it into Blender (or you 3D Application of choice), get a screenshot from Google Maps of the area where you are going to project (or any map that may be helpful), get in Top Orthogonal view and place the 3D model where it is in reality and the camera where the projector is going to be placed.
Try to calculate the angle of projection of the projector and apply it to the virtual camera.
In my last projections we had also to use a laser tool to get the proper height of the prjector in respect of the facade, here the terrain is usually very steep, in a flat land it should be easier.
Ok now you have recreated a virtualization of the real scene, if you try to project a screen shot from the Blender render it won t fit anyway, but read on…

Here is where vvvv steps in, get a wireframe render from blender (let s call it “the mask”) and load it in vvvv, project it over the facade and adjust it with an Homography node so that it fits nicely, save and go home.
Now you are ready to get wild, do some animation in Blender, prepare a Box2D interactive vvvv patch based on a front orthogonal view of the facade or a 3D owrk based on the Projector node, whatever you do all you need to be sure about is that it will fit in the mask, keep the mask on the final render and blend it with your work to check out how it works out.
You can use an Homography node for every scene of your projection, so they all fit within the mask.

Time to show off ? Now that everything is centered on the mask all you need is to project the mask (you could reduce it just to 4 pinpoints so nobody will notice you fixing the projection) and use th efinal Homography node for the final tweak; the projector has been moved from the original position? No problems, you can fix it in a sec.

DISCLAIMER! I haven t tested this setup in a real 3D projection, a facade although being a 3D element, isn t usually very deep, so the amount of distorsion that may cause manipulating the final render in 2D may not be noticed.

Ideas, comments, hints ?

wow thanks for sharing!!!
i didn’t know the existance of such a powerful tool to create 3d from still pictures!

grazie simone and please keep up the thread if u have any news!
ps.
if it’s not a problem maybe u can share a photo of the facade you did and a photo of the 3d mesh that the software produced…or any other visual representation of the procedure u’ve described above :)
any tutorial would be obviously very appreciated by me

this whole homography tech is fine if the building is flat. Once you start doing buildings that aren’t flat and using multiple projectors you’ll start having problems. If the building is flat then all you really need is a good photo from about where your spectators will view from.

The 3D models from pictures is a great as a reference but I’ve found that often that large flat faces dont come out that square. There’s always tweaking to do but you never really know which point was the correct one. The best way as is build from scratch using building plans, architecture drawings and land surveys. You may even find a 3D model/cad drawings exist if you contact the right people. Well worth the effort if you can get all this data and get an accurate model from the offset.

Anyone know?
http://www.photosculpt.net/videos/

Agree on the first sentence, but I don t on the second one, since there are 3 factors to take into account: projector(s)‘s position, spectators’ (if any) position, and orthographic view. If I had to choose one I d go for the projector’s position if there is one projector, or for the middle point between the 2 projectors if 2 are used (which would probably be the same as spectators’ position.

Now, why bothering with photo modeling software to get a perfect orthographic view? basically because of the weird position of the projectors and spectator I has to deal with lately.
The facade was placed at a strong angle both horizontally and vertically in respect to projector and public, and it was impossible to get a proper front faced image of it (think living on a highly populated volcanic island with a 4000 meters high volcano rising from the coast line), so photo modeling was the only way.
Now I am going to project on a 4 meters high, 15 meters wide pile of broken fridges, once again, there is no way to get a picture from a proper distance, so the only choice is to reconstruct the scene with photo modeling software and get a orthographic shot within Blender.

yep, but here in banana republic it s not so common to get any of that… better get used to live without atm

I am just beginning with all of this so I am trying to gather as much experience and points of view (never better said) as possible. I ve spent much more time to find out about the various technique about projection than creating the final content which is lame, but entertaining, after all, so I think i keep this post updated for anybody who wants to join in the discussion (it s not possible I am the only one living in a volcanic island…)

cheers
Simone

Attached the virtual scene of the facade projection.

thx 4 sharing and for keeping the thread up simone!!

…to build models from pictures (for those without paid-for 3d software), i can recomend a combination of ‘free-cad’ and ‘blender’.

‘free-cad’ (an alternative to autocad), as it is real easy to put in pictures, resize, etc. and then all the lines just go snap, snap, snap.

http://sourceforge.net/apps/mediawiki/free-cad/index.php?title=Main_Page