Minimizing support structures is crucial in reducing 3D printing material and time. Partition-based methods are efficient means in realizing this objective. Although some algorithms exist for support-free fabrication of solid models, no algorithm ever considers the problem of support-free fabrication for shell models (i.e., hollowed meshes). In this paper, we present a skeleton-based algorithm for partitioning a 3D surface model into the least number of parts for 3D printing without using any support structure. To achieve support-free fabrication while minimizing the effect of the seams and cracks that are inevitably induced by the partition, which affect the aesthetics and strength of the final assembled surface, we put forward an optimization system with the minimization of the number of partitions and the total length of the cuts, under the constraints of support-free printing angle. Our approach is particularly tailored for shell models, and it can be applicable to solid models as well. We first rigorously show that the optimization problem is NP-hard and then propose a stochastic method to find an optimal solution to the objectives. We propose a polynomial-time algorithm for a special case when the skeleton graph satisfies the requirement that the number of partitioned parts and the degree of each node are bounded by a small constant. We evaluate our partition method on a number of 3D models and validate our method by 3D printing experiments.

No abstract

We introduce SketchGNN , a convolutional graph neural network for semantic segmentation and labeling of freehand vector sketches. We treat an input stroke-based sketch as a graph with nodes representing the sampled points along input strokes and edges encoding the stroke structure information. To predict the per-node labels, our SketchGNN uses graph convolution and a static-dynamic branching network architecture to extract the features at three levels, i.e., point-level, stroke-level, and sketch-level. SketchGNN significantly improves the accuracy of the state-of-the-art methods for semantic sketch segmentation (by 11.2% in the pixel-based metric and 18.2% in the component-based metric over a large-scale challenging SPG dataset) and has magnitudes fewer parameters than both image-based and sequence-based methods.

Reducing the volume of support structures is a critical means for saving materials and budgets of additive manufacturing, and tree structure is an effective topology for this purpose. Although a few articles in literature and commercial software have been devoted to developing tree-supports, those tree-supports are generated based on geometry optimization or user-defined parameters, which cannot guarantee a minimum volume with robust fabrication guarantee. To address this issue, we propose a set of formulas for stably growing the tree-supports with physical constraints based on 3D printing experiments using fused decomposition modelling (FDM) machines, and a volume minimization mechanism using a hybrid of particle swarm optimization (PSO) method and a greedy algorithm. We show that this combination is effective in reducing the volume of tree-supports and the simulations reveal that the volume curves monotonically descent to a constant within a short time, and our experimental results show that the models with the tree-supports can be manufactured stably.

We introduce SketchGCN, a graph convolutional neural network for semantic segmentation and labeling of freehand sketches. We treat an input sketch as a 2D point set, and encode the stroke structure information into graph node/edge representations. To predict the per-point labels, our SketchGCN uses graph convolution and a globallocal branching network architecture to extract both intrastroke and inter-stroke features. SketchGCN significantly improves the accuracy of the state-of-the-art methods for semantic sketch segmentation (by 11.4% in the pixel-based metric and 18.2% in the component-based metric over a large-scale challenging SPG dataset) and has magnitudes fewer parameters than both image-based and sequencebased methods.