Canyon Scene Reconstruction: Generating a 3 Dimensional Mesh from Video Footage

Photo of Rick Barrett

Instructor Rick Barrett

Share this video
  • Duration: 05:58
  • Views: 2908
  • Made with Release: 19
  • Works with Release: 19 and greater

Create a 3D mesh using C4D’s Scene Reconstruction.

In this video, we’ll explore how the Area Weight and Photometric Weight affect the 3D mesh generated by Cinema 4D’s Scene Reconstruction. After finding suitable values, we’ll smooth out the resulting mesh via a Phong Tag and Smoothing Deformer.



In this video, we're going to generate a 3D mesh based on reconstructed points from our video footage. At the end of the last video, we had generated this point cloud based on the medium quality settings or the medium quality preset within the reconstruction tab. The only difference is that I activated this Point Cloud Only option in order to only generate the point clouds so that we can focus on meshing within this tutorial. Now, because we've already run a scene reconstruction, we can generate the mesh without running the Reconstruction again. We can adjust the Area Weight and Photometric Weight values and play with these and regenerate the mesh and we only have to rerun the full reconstruction if we want to adjust the values up in this area. So, I've already run a few examples of the meshing results with different Area Weight and Photometric Weight values in order to show you how those affect the resulting mesh. Basically, a higher Area Weight value is going to only give you mesh in areas that have higher confidence. This is sort of like a confidence threshold. So, if you set the Area Weight to 10%, as it is by default, you get a mesh that looks something like this, which is really, fairly complete. It breaks up a little bit here towards this back bluff. That is the result that we get with a 10% Area weight. If we increase that to 35% Area Weight, you can see that we lose mesh because it's got less confidence in those areas back here. And if we go all the way up to 100% Area Weight, you can see that we're not reconstructing much at all at this point. We're really just getting that very front bluff or escarpment. So typically, we're going to want to use very low values for this. In fact, let's look at what happens if we set the value all the way down to 0. Here, we've got a fairly complete mesh without many holes. But what we get is some areas that become connected that probably shouldn't be. I mean, if we go back and look at the footage itself, you can see that this area here is more canyon, that should have more detail, and we don't necessarily want to create a giant backstop at the back of the footage that we get with an area weight of 0. So, small values but not too small. Now, the difference between Area Weight and Photometric Weight is that Area Weight is more of a geometrical calculation whereas Photometric Weight looks at the image itself. And so, Photometric Weight can be a little bit more accurate. Let's look at 5% Area Weight and 0% Photometric Weight versus 0% Area Weight and 5% Photometric Weight. It's very close, but I think the Photometric Weight is slightly better here. You can, of course, combine these and this is the result when we use 5% in both values. This is what we get if we go down to 2 for both, and you can see that that's completing more and more of the mesh here. And we've still kept this backstop out of the back of the canyon. Now, you might get to a point where you decide that you do need to actually reconstruct the scene again, and that's the point that I got to. I decided that I wasn't quite happy with the level of gapping that's happening, especially back here on this bluff, and I wanted to see if I could close that up a little bit. So, I went back to the motion tracker in the reconstruction tab, and I decreased the minimum texture detail because I figured that that might be what was the limiting factor in this case. Because the texture is not incredibly detailed here, it's a little bit blurry, so, I dropped this down to .1 from .5. And I also went ahead and ran three iterations. And once I generated the mesh based on that, I got a result that looked like this. And it's a little bit hard to tell, but we are getting a more solid mesh than we had before. Once you're happy with the mesh that's generated, there is a number of things that you can do to smooth it out. Obviously, this is a very rough mesh, very highly polygonalized and it's going to need a little bit of refinement in order to do much with it in Cinema 4D. So, the first thing that you can do to just add some simple smoothing to it, is add a Phong tag, and you'll find that within the Cinema 4D Tags menu. And once we do that, and I'll turn off the lines, you can see that that smooths out the appearance of this quite a bit. Now, it's still pretty rough and so what I actually prefer to do is to add a Smoothing Deformer. If you hold down the Shift key while you create the Smoothing Deformer, it will automatically be added as a child of the mesh, and that way it's already applied and you can see the deformation immediately. Now, we can adjust the strength of the smoothing here and decide how much of the original roughness we want to maintain. But I'm happy really, actually, with going all the way to the 100% smooth. And you can see that now we've got a mesh that I think we can work with pretty well. Now, you might still have some holes that you'll need to try to model around, but this is giving us a good starting point from which we can build some of the effects that we want to create with this reconstructed mesh.
Resume Auto-Scroll?