Like just about everything else in today’s world, knitted textiles and clothing are produced using automated machinery. Computerized knitting machines can rapidly churn out garments and other textile goods, and can then be reprogrammed for the next job. That efficient mass production is the key to making affordable textiles. But actually programming computerized knitting machines isn’t a simple task, and it currently requires a great deal of expertise. That’s why researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed new knitting design software that anyone can use.
This software, which is like a conventional CAD (Computer-Aided Design) program for automated knitting machines, is actually the result of two projects at CSAIL. The first is call InverseKnit, which is a system for reverse-engineering a photo of a knitted garment into instructions for computerized knitting machines. InverseKnit relies on machine learning, and was trained on a data set that included many images of knitted items and their instructions. When presented with an image of an unknown garment, it can automatically generate new instructions that are accurate 94% of the time. That could result in massive savings for companies that commonly reproduce existing designs.
But the researchers also wanted a way for inexperienced users to modify and customize their own designs. That’s why they developed the CADKnit software, which allows non-expert users to produce new designs by modifying templates. They can, for example, start with a template for a basic glove. By using simple photo editor-like tools, they can modify the knitted patterns. CADKnit even has the ability to use take illustrations and make them part of the design. Once the design is complete, instructions for computerized knitting machines are automatically generated. In tests, subjects with no previous experience were able to easily use the software to create their own designs.