TL;DR
This paper introduces PushPull++, a novel tool for rapid polygonal modeling that significantly improves upon existing "push-pull" techniques. It addresses key challenges in local mesh editing, such as adaptive face insertion, robust adjacent face updates, and handling edge collapses, leading to a more intuitive and efficient user experience.
What Problem Does This Paper Address?
Commercial 3D modeling software often includes "push-pull" tools that allow users to intuitively transform faces, edges, or vertices of a polygonal mesh and have the surrounding geometry adapt locally. However, existing tools have several limitations: they may only work with orthogonal faces, fail to add new details, or struggle with self-intersections and edge collapses, especially in complex scenarios with non-planar or slanted surfaces. This makes detailed and precise editing of man-made objects, common in architecture and design, cumbersome and time-consuming. The paper aims to overcome these limitations by providing a more robust and versatile push-pull operation.
Methods
The core of PushPull++ is a novel mesh modification method that operates on arbitrary polygonal meshes without requiring watertight models or constructive solid geometry. The authors introduce three key contributions:
- Adaptive Face Insertion: This method intelligently decides whether to insert new faces or adjust existing ones based on angular thresholds between adjacent face planes and the target plane. This allows for adding details only where needed.
- Generalized Mesh Update Algorithm: This algorithm efficiently computes new vertex positions and updates the local mesh topology. It handles cases where a single vertex needs to be replaced by multiple new vertices to maintain planarity of adjacent faces, even when vertex valences are greater than three.
- Stepwise Modification for Intersection Handling: To prevent self-intersections and edge collapses, the tool employs a stepwise approach. It iteratively identifies conflict events (like shrinking edges) and applies modifications with intermediate planes until the target plane is reached. This avoids the need for complex boolean operations.
Additionally, the paper describes a user interface that automatically proposes useful drag directions, enhancing intuitiveness and reducing user effort.
Datasets
Not discussed in the paper. The paper focuses on a novel modeling tool and its evaluation through user interaction counts and qualitative examples.
Key Insights and Ideas
- Adaptive face insertion is crucial: Simply always inserting new faces or always adjusting existing ones leads to limitations. An intelligent decision-making process based on geometric properties is key.
- Maintaining planarity is complex: When a face is moved, updating adjacent faces to remain planar, especially with high-valence vertices, requires sophisticated geometric calculations.
- Stepwise modification elegantly handles degeneracies: Instead of relying on computationally expensive boolean operations for edge collapses and self-intersections, a step-by-step approach with intermediate planes offers a robust and efficient solution.
- User interface design matters: Automatically suggesting useful editing directions significantly reduces the cognitive load on the user and speeds up the modeling process.
Key Results
The paper demonstrates that PushPull++ significantly outperforms existing tools like AutoCAD, SketchUp, and Maya in terms of efficiency. For complex modeling tasks, PushPull++ requires up to an order of magnitude fewer mouse clicks and geometry-changing interactions. For instance, in one comparison, PushPull++ required 1 interaction for certain steps where AutoCAD needed 24, SketchUp 26, and Maya 8. The method is also shown to be performant, achieving at least 30 updates per second on mainstream hardware and scaling linearly with the number of affected faces.
Limitations and Future Work
The authors acknowledge that PushPull++ currently only prevents local self-intersections. They suggest future work could extend the algorithms for global intersection handling, which presents an interesting challenge in crafting intuitive user responses. Additionally, extending the approach to handle curved surfaces is mentioned as a promising direction.
Summary updated