New in Cinema 4D R19: Recreate your Filming Location with Scene Reconstruction

Photo of Cineversity

Instructor Cineversity

Share this video
  • Duration: 07:16
  • Views: 20658
  • Made with Release: 19
  • Works with Release: 19 and greater

Create a model of your filming location using the Scene Reconstruction Tab of the Motion Tracker object.

Cinema 4D Release 19 further improves the Motion Tracker system by adding the ability to generate 3D geometry from your motion tracking points. This can be used to create reference geometry for modeling, HUD effects, collision geometry, or even a quick-and-dirty shadow catching object.

Learn how to take a piece of footage that has already been camera tracked and create a model of your filming location using the Scene Reconstruction Tab of the Motion Tracker object.

Topics Covered Include

  • How footage should be prepared for Scene Reconstruction

  • What sorts of shots and objects will result in the best reconstruction

  • How to setup an object to be used as a compositing background

  • How to Project Video Footage back onto your Reconstructed Mesh

  • Rendering the Vertex Colors on your Object

show less

Transcript

In this Cinema 4D Release 19 Quick Start, I'm going to show you how to use the new reconstruction tab in the motion tracker object to create 3D geometry from this source footage. Before you can reconstruct the geometry of your environment, you have to first track your footage. And I accomplish that in this scene by using the full solve method. So, it was basically a one click solution to get my initial 2D tracking points, and even to generate all of these 3D nulls that you see in my scene. Now, after I generated those nulls, I went in and I adjusted the orientation of my scene with this planer constraint. I positioned it at the world origin with this position constraint, and I even adjusted the scale with this vector constraint. And I did all of this so that if I was to add a figure object to my scene and press play, it would be appropriately sized and more or less locked in place. Now, if you add an object to your scene and you find that it is slipping quite a bit, that means you need to go back and get a better camera track before you attempt reconstructing your scene. So, I now have this good track and I'm going to go to the reconstruction tab here. And it's basically a one click solution. You choose the level of quality that you want and choose run scene reconstruction. In the lower left hand corner, you'll get a status update of what's going on, and sometimes you might see that some of these steps are repeating, and that repetition is controlled here by the iteration setting. For the preview though, we've only got one iteration, so this should be relatively quick, at least by the standards of scene reconstruction. All right, our preview has finished and what we get is a scene point cloud object. This is a polygon object that has no polygons, it just has points. And each of these points has a color assigned through this vertex color tab. So, if I playback through my scene, we will see, well, at least from your perspective, probably not much. The video compression is probably going to crush this pretty brutally. So what I'm going to do is hide my auto features and I'm going to pop out of my soft camera so we can look at these reconstructed points. And you'll see that these more or less look like the nulls we that we had previously, but they've got some color applied. Now, this indicates to me that we are going to be able to get a solve period. And that's really the main thing you're doing when you use the preview mode. You're just trying to establish that you can create any geometry whatsoever. And that's an important step because some footage is just not going to reconstruct well. Generally because there's not enough texture on your surfaces or lighting is too difficult or it just can't track what's happening in between frames. So, what I want to do now is see what this might look like as a mesh. And this is a good way of working. So, what you want to do is first generate, often in point cloud only, to figure out if you're getting a good density of points and once you've done that, choose generate mesh. And that's going to take a few moments to go back in and calculate how all of these pieces should fit together. And what you'll get back is, at least with the preview mode, something like this. And we see that this tree is really successful in being created and that's because it's got a lot of really high frequency detail in the bark here. It's got very clear but very distinct patterns that are going to be easy to recognize from multiple angles. So, now that we've got a sense that this is going to work, what I want to do is change my preset here from preview to medium quality and then I'm going to click on generate mesh. All right, now that that calculation has completed, we can play back our footage and see how well our scene has been reconstructed. It seems like our geometry is tracking quite well with our footage, no significant slippage. I think this will be good for us to work with. Now, our display is just a little bit messy, so I'm going to ALT or option-click to hide my auto features here and I'm going to hide my point cloud as well. And what are some ways that we can now render this geometry? Well, probably the simplest way is to take your motion tracker, go to the footage tab and choose create background object. What that's going to do is create a background object and apply a material to it that's got a frontal projection. I can now duplicate this material and place it on my scene mesh. And when I play back, you'll see that our footage is being live-projected onto our geometry. And it looks, more or less, like we want. If you want, you can go in there and add in a smoothing deformer, in order to soften out some edges. And if you were to add in a compositing tag, it would make this a compositing background, you'd have a good starting point for shadow catching. Now, what if you want to render this without projecting your footage back onto it? Well, for that, we have this vertex color tag which allows us to see the color that has been stored at each and every vertex, but this does not render by default. So what we have to is create a material that will hold that vertex tag. So, in my color channel, I'm going to go to texture and create an effect vertex map shader. And from here, I'm going to drag in my vertex color tag, drag this onto my mesh. And I guess I can call this "vertex color." So, now when I render, we're going to see a color applied to each and every vertex. Now I am seeing a little bit of specular highlight which I will kill and re-render here. And again, we are seeing this vertex color. Now, this is not as detailed as projecting your video back onto the surface of your object and that's because it's just storing a vertex color. It's not storing texture information. And that's the process for converting your footage into geometry. You can use this directly for simulating LIDAR scans, break it up for HUD effects, smooth it out for collision geometry, and even use it as a reference for more accurate polygon modeling. If you enjoyed this video, you may want to check out the rest of our "New in Cinema 4D Release 19" series on cineversity.com, or you can subscribe to hear more about new videos as we release them.
Resume Auto-Scroll?