I've been using the Three a bit recently and I've found it actually captures some black objects, which I was not expecting. Due to how cranked all the exposure and brightness settings are, there is a bit of random 'noise' data which is floating off the model. For most cases I can just use the lasso tool to select the points, but for some models it's very difficult to selecting portions of the model.
Maybe there can be a feature similar to what you'd see in Revopoint's software to automatically select floating points?
Also, sometimes I don't mind meshing a pointcloud with a bit of missing data (just lookinga t one part of it, etc.), but the software tries to close all these holes, leaving a gnarly looking mesh. I find the hole closing on the Matter and Form software exceptionally good most of the time, but in the cases we don't care if the model is watertight, is it possible to dissable it?
Thanks!
Hi Justin!
Thanks for the feedback. These are some great suggestions that I've passed on to the dev team.
We hadn't really considered how to do deal with stray points that are mostly enclosed; these are difficult to remove with the lasso without accidentally removing the data behind them, and there isn't an easy to access angle where they are entirely isolated from the rest of the point cloud. We'll try to brainstorm some ways to address this specific situation in an easer and more intuitive way.
In regards to meshing objects with missing areas; our meshing algorithm prioritizes watertight meshes, and this can lead to some ugly artifacts if too much of the object is missing. While it's obvious to our brains how to fill in missing sections (because we know what the object looks like) the computer does it algorithmically; sometimes this works well and sometimes it doesn't.
There's two ways to proceed here. The most obvious is to do more scans to fill in more of the gaps. The less missing stuff, the better the software will be at guessing the missing portions. The other alternative is to export your point cloud and do the meshing in a program that has more robust meshing options; something like MeshLab. Here you can pick from a variety of meshing algorithms that have different priorities and advantages. Our meshing is very "generalist" and works best for near-complete point clouds that are destined for printing, but there are definitely other use cases where it is less ideal. We may explore giving more control over this part of the software in the future, but in the interim you may want to experiment with with the ball pivoting algorithm in MeshLab.
