We will describe AmiGo, an approach for generating crochet instructions (patterns) from an input 3D model. We focus on Amigurumi, which are knitted stuffed toys. Given a closed triangle mesh, and a single point specified by the user, we generate crochet instructions, which when knitted and stuffed result in a toy similar to the input geometry. Our approach relies on constructing the geometry and connectivity of a Crochet Graph, based on the geodesic distances from the seed point, which is then translated into a crochet pattern. Using the topological structure of the geodesic distance function, we segment the shape automatically into chrochetable components, which are connected using the join-as-you-go method. We relate the shapes' curvature functions to crochetability, show how non-crochetable shapes can be fixed, and demonstrate the applicability of our method to a large variety of shapes and geometries. We will additionally discuss ongoing research directions and future work.

The talk will be in Hebrew.