KnitGIST: A Programming Synthesis Toolkit for Generating Functional Machine-Knitting Textures
summary
KnitGIST is a program-synthesis pipeline that generates machine-knittable KnitSpeak programs from high-level objectives by mapping weighted objectives (curl, shrinkage, stretch, opacity, style, symmetry, imagery) to small tactics (flip/lean stitches, mirror, replace, swap). It scores N-best random textures, computes an expected-improvement (EIS) for tactics, and iteratively modifies candidates; a 1,500-pattern corpus classification reaches 95.6% accuracy (κ=.86). Demo builds a stretchy, curling lampshade with light-transmitting stripes and cabled bands.
AI-written summary · about
related talks
17mKnitPicking Textures: Programming and Modifying Complex Knitted Textures for Machine and Hand ...
41mLanguages for 3D Industrial Knitting
28mExploring Self-Embedded Knitting Programs with Twine
18mKnitting Skeletons: A Computer-Aided Design Tool for Shaping and Patterning of Knitted Garments
26mKnitting for Javascripters
29mA Foolish Quest: Creating Knitting Patterns Using JavaScript
27mProgramming with Yarn
25mpyKnit: math tools for knitters