|
|
|
#1
17th February 2011, 10:13 AM
|
|||
|
|||
3d Projection mapping
Hi friends..
i am newbie in this field. I have been approached by a company to do something like the link below for there car launch. Being a visual effects artist compositor i understand the 2d projection system But to do in real 3d i have no experience. i wud like to ask the seniors here what shud my prep be to achieve such stuff. wht are the basic question i shud adhere to and keep in mind Projection res The cgi renders res the maths evolved thanks in advance.. 
|
|
#2
17th February 2011, 10:58 AM
|
||||
|
||||
|
Well good luck,
Personally I'd say that map is far too complex to provide a 'do this, then this' answer. I would say that every map requires its own solutions, so explaining what they did above is not the same as explaining how to map what you are trying to achieve, so you should be provided useful information, whats the space, what is the target object, (a car requires a completely different setup to a large building facade say). Then you need to know what 3d package you are planning on using (even deciding what 3d package you need is subject to a whole host of questions as each has different advantages) and finally what type of content your trying to create. If you have no experience in 3d then I'd honestly say that in any commercial time frame you have zero chance of learning any 3d program well enough to achieve results anywhere near this level, getting past the basics in most 3d packages often takes 6 months of brainwork, let alone the time needed to fully understand the various mapping techniques, advanced projection alignment, and syncing the video across all the various beamers.... something that has taken many months or years to learn, and even then few understand the entire work flow. Mapping is incredibly time consuming and very complex to get right if your aiming at anything beyond simple shapes. To do anything like the mapping on that car is a serious chunk of pretty advanced work. To even model something like the car above accurately in 3d and align to it correctly to a real object takes a huge learning curve, especially given you have no working experience. Its especially difficult if your working with organic shapes, trying to create content that works in the right perspective, follows the lines of the shape, aligns correct over all various projectors and makes sense from the viewers point of view, all relies on very high accuracy and top notch 3d production skills. To achieve any type of mapping you need to understand the enormous reliance on the final environment, you need to understand where your projectors sit, what lenses are required for the throw ratio and desired brightness, how many beamers you need, and what coverage each one achieves, you need to know what the mapped object is, how to recreate it, what angles will the viewers be able to see. If you manage to learn the requirements it still doesn't mean your map will be great, understanding how to enhance the objects 3d form, and to create content that works for the viewers angle, and most importantly look believable, takes massive amounts of time and experience... I won't even go into the full complexity of 3d content creation, even deciding on res depends on playback method, number of beamers, their angles, beamer resolution, what area your trying to cover and what shape your trying to fit. Also if your doing the map to create an advert rather than a live performance then you need to think about how to film the map to achieve a quality look so that may also involve additional lighting and ensuring against issues like playing back interlaced footage. Basically I'd advise you to work with a company that has experience in the field, maybe put a call out for someone to work with. To simplify my method I use cinema4d, taking real photos from the various projection points, making sure they are taken on the correct lens so I can mimic the camera in cinema4d, I'd then recreate the mapping target as a 3d object, that could be done by using a 3D scanner to create a point cloud of the geometric samples, or recreation from photos using maybe c4d's projection man tool or Autodesk imagemodeller, even maybe corner pin tracking data from something like pftrack. I'd then create some content using a combination of mograph, compositing and c4d, then map that content back onto the 3d objects in cinema4d possibly via bodypaint to allow a more artistic texture painted use of the content, and finally render out all the cameras for more post compositing. I'd then process playback either via vvvv for very large complex mapping, or for simple stuff using the IRmapio plugin for resolume and do the final alignment via decent projectors that have quality alignment tools like those provided by barco or christie. I warn you though even one of those stages relies on a whole host of applied knowledge, so one stage like applying your content accurately in 3d relies on programs like c4d's bodypaint or projection man and a firm understanding of the settings and types of texture map, how they effect render time, quality, and texture resolutions, and even use of animated textures... It all depends on what your trying to create, so even learning to texture map in 3d will take you months to understand and start getting right, same for rendering, same for modeling, same for creating complex animation in mograph, same for understanding how to work out lenses, projection stacking, and advanced alignment. Advanced mapping is about joining many disciplines, when each on its own can take years to achieve a consistently professional output. So start small, practice on small easy shapes, start learning 3d, try and recreate an object and align it to a photo... You may find you get stuck doing something basic, so find your limits first, then design a project that's achievable. Last edited by evomedia; 17th February 2011 at 01:39 PM.
|
|
#3
17th February 2011, 02:06 PM
|
||||
|
||||
|
A car is very complex, it has organic shapes (not flat). This would make it really hard to map.
The easy approach: -split the car up in to easy parts (like a door, the hood,...) and measure them perfect. -now you can make with your 3d program animations for each part. Add all the parts to 1 big composition (make sure it's synced). -Use IR Mapio to map this stuff on the car. 1 tip: TEST it a lot!!! when you have your first "framework" of the car, try to project it on the car. Then do your animation  (a bit offtopic  Quote:
|
|
#4
17th February 2011, 02:22 PM
|
||||
|
||||
|
Quote:
If you want to create a 3d model using just photos then check out autodesk imagemodeller that I also mentioned, or projection man in c4d. more about modelling from tracking point cloud data http://www.photogrammetry.ethz.ch/ge...p_modeling.pdf If it was a simple object you could use a few track points to create corner pin data that can be used to in ae Last edited by evomedia; 17th February 2011 at 02:52 PM.
|
|
#5
17th February 2011, 02:25 PM
|
||||
|
||||
|
I don't know pfdata. Will try it if it has a demo
Last edited by evomedia; 17th February 2011 at 02:59 PM.
|
|
#6
17th February 2011, 02:59 PM
|
||||
|
||||
|
Quote:
http://www.pfhoe.com/ Another program for converting point clouds into a mesh is sycodes point cloud, which just basically uses the point cloud as a base to drap a mesh over. http://www.sycode.com/products/point_cloud/index.htm If your in cinema 4d then getting the cloud points into a useable format is the main issue, you need to turn on axis snapping if you're trying to snap to nulls (point data). You might first need to select all the nulls and run the Axis Center function. To create a mesh in c4d sometimes using metaballs is a useful method, the same process as taking realflow simulation data and meshing that. http://www.molecularmovies.com/image...Dmetaballs.pdf I still think that autodesks imagemodeller provide a great solution, as you use the photos to model by setting the same point in different pics which it then uses to work out the 3d form to speed up modelling. Last edited by evomedia; 17th February 2011 at 03:40 PM.
|
|
#7
17th February 2011, 03:25 PM
|
||||
|
||||
|
Thanks for the great info, it looks like a very good technique for 3d mapping.
|
|
#8
17th February 2011, 03:45 PM
|
||||
|
||||
|
While I remember also meshlab is useful when working with point cloud data and remeshing... plus its free
http://meshlab.sourceforge.net/ Last edited by evomedia; 17th February 2011 at 03:53 PM.
|
|
#9
17th February 2011, 05:57 PM
|
||||
|
||||
|
back to the thread topic, can't believe I didn't see this link on the page... a making of video.
It does indeed show how complex a task this is, but good viewing, pretty much outlines the same process I described Last edited by evomedia; 17th February 2011 at 06:04 PM.
|
|
#10
29th November 2011, 04:22 PM
|
|||
|
|||
3D Projection mapping for a car...
In the context of the posts above, I have a related query.I too am trying to figure out the way I should do this. The position of my car is almost the same as this video
http://www.madmapper.com/2011/11/14/...n-car-mapping/ If I have a replica 3D max model of my car, and plan to work in max and aftereffects for the animation portion, how do I proceed? I have looked up VP6 and am also considering mad mapper for the projection, will these do the trick? How do I map specific areas accurately, especially of curved surfaces. Do i need multiple projectors for the different faces and how do I rig the projectors? As you can see, I'm an absolute novice at this, so if anyone could point me in the right direction, I would be eternally grateful. So many questions...too little time. 
|
 |
| Thread Tools | |
| Display Modes | |
|
|
Linear Mode
