I’ve been using Quad Remesher in Rhino for a while and have identified several areas where the tool could be significantly improved. I’d like to share my observations and gather feedback from the community and developers to enhance its functionality.
Issues Observed
- Detail Recognition and Preservation
- Current Performance: The tool appears to lack the capability to accurately recognize and preserve high levels of detail within the mesh. For instance, in the images provided (Pic 1 and 2), there is a closed polysurface measuring 90 cm in length, with filleted edges and two convex rings (ring diameters vary from 1 cm to 0.1 cm). When setting the Target Quad Count to 4000, the resulting mesh comprises 5730 faces, both with and without SubD mode enabled. As evidenced in the images (Pic 3, 4, 5, 6), the mesh fails to retain critical details and exhibits shading and topology issues, particularly along smooth boundaries.
- High Target Quad Count: Even with the Target Quad Count increased to 100,000—an excessively high density for such a simple mesh—the same issues of detail loss and shading inaccuracies persist (Pic 7). Utilizing a mesh with 100K faces for a single object is impractical for rendering and presentation, especially in scenes containing multiple such objects.
- Uniform Subdivision Algorithm
- Current Performance: The core issue is that Quad Remesher subdivides the entire object uniformly, regardless of the set target quad count (4000 or 100,000). This approach fails to adapt to the detailed parts of the mesh and applies the same subdivision method to both intricate and simple regions alike, leading to inefficient use of polygons and computational resources.
Key Areas for Improvement
- Adaptive Detail Recognition
- Suggestion: Quad Remesher should implement an adaptive algorithm capable of discerning and preserving critical features such as sharp edges, small details, and topographical variations. The current indiscriminate subdivision of the entire mesh results in unnecessary computational overhead and RAM usage.
- Efficient Mesh Density Allocation
- Suggestion: The algorithm should allocate mesh density intelligently, concentrating polygons in areas with high detail while minimizing tessellation in flatter, less detailed regions. This would ensure that the final mesh retains essential details without excessive and redundant subdivisions.
- Shading and Topology Enhancement
- Suggestion: Improvements in the shading and topology algorithms are necessary to produce cleaner, more accurate meshes. The existing method struggles with maintaining smooth boundaries and correct shading, leading to a jagged and visually unappealing output.
Visual Examples
To illustrate these issues more clearly, I’ve included several images showing different scenarios where Quad Remesher’s performance could be improved:
- Pic 1 & 2: The initial closed polysurface with filleted edges and convex rings.
- Pic 3, 4, 5, 6: The mesh generated with a Target Quad Count of 4000, highlighting significant loss of detail and shading issues.
- Pic 7: The mesh generated with a Target Quad Count of 100,000, still displaying detail loss and shading problems.
Conclusion
While Quad Remesher is a powerful tool, there are substantial areas for enhancement to improve its effectiveness. By focusing on adaptive detail recognition, efficient mesh density allocation, and improved shading and topology algorithms, we can significantly enhance its performance and utility. The current uniform subdivision approach must be re-evaluated to better react to mesh details. I look forward to your thoughts, feedback, and any additional suggestions.
