


And there is a really detailed tutorial on how to do that. Next we use Blender to overlay a human skeleton-like armature over our mesh. This is not ideal, I need to think about how I can block the vertex movement to prevent this without losing too much performance. File > Export Mesh As… *.ply format (or anything else Blender supports) on on edited On moderate simplification ratios, simplifier creates overlapping faces, which causes MeshLab to generate incoherent normals.And then use Quadric Edge Collapse Decimation with Simplify only selected faces checked. So here is what you can do: on the MeshLab toolbox panel select the tool called Select faces in a rectangular regions and select all, but the face. The last checkbox proved to be very useful: when you simplify you most probably would not like to simplify the face very much. I used Filters > Remeshing, Simplification and Reconstruction > Quadric Edge Collapse Decimation: After that it is makes sense to simplify your model, you will not notice the difference visually and Blender will be much happier to work with smaller mesh.I used Filters > Point Set > Surface Reconstruction: Poisson which seems to be the most poplar way to do it (it will take couple of minutes). Now we are finally ready to perform the surface reconstruction.Render > Show Vertex Normals allows you to see the normals, here they are:.So we invert them Filters > Normals, Curvatures and Orientation > Per Vertex Normal Function But we want our normals to look out, not in.Which resulted in (almost) all normals looking at the point (0, 0, 0) as we requested. First step was Filters > Point set > Compute normals with the following parameters.This operation is called surface reconstruction and can be easily done for example using MeshLab software: Triangles are better than points because once you have them you can compute lighting effects and give volume to your model. It all started in a device similar to this one: The scanner packed me into a point cloud with ~400,000 points, which looks like this:įirst we need to convert point cloud into a mesh: a graphical model, which consists of triangles. And today, to avoid doing all the things I really needed to do, I decided to go through the pipeline from a 3D scan to an animated model. So it happened that yesterday I did a 3D scan of myself.
