Open JYChen18 opened 3 weeks ago
Hi When the model has sharp corners or thin plates and the sampling is sparse, the dual of RVD may appear non-manifold or erroneous. This is one of the defects of RVD itself. We cannot guarantee the manifold mesh output, which is also one of the limitations of CWF. As for the division by zero error you encountered, I don't think I have seen it before. Maybe you can try a different model or sampling.
Best, Rui
Thank you for your quick reply!
After sampling 1000 or 2000 initial points, the runtime error is resolved, but the non-manifold issue still persists.
Finally, I found another repository, ACVD, which is capable of generating a manifold output. While it doesn't perform as well at preserving weaker features compared to your approach, it runs significantly faster and suits my current needs.
Since I’m not familiar with this area of research, I'm unsure if their method for enforcing manifold outputs could be applied here. For now, I plan to use their method to save effort. However, I’d be happy to revisit your excellent work if there is any update addressing the non-manifold issue.
Thanks again for your work and help!
Thanks for your suggestion! We will consider that!
When I try to simplify my mesh test.zip, I get the runtime error of "Division by 0" around the 47th iteration. The complete output is
Another problem is that the resulting mesh is not strictly a 2-manifold, which is necessary for my downstream task, i.e. SDF query.