Skip to main content

MIT Researchers Use Machine Learning to Make 3D Knitting a Cinch

Computerized knitting machines have proved a boon for the knitwear industry.

Systems like Nike’s Flyknit and Shima Seiki’s Wholegarment are designed to minimize labor, curtail production time and pare back waste. Even brands like Ministry of Supply are banking on 3D knitting to promote bespoke clothing made on the fly. But programming these machines requires skill and patience, not to mention a healthy awareness that a single error can wreck an entire garment.

To streamline the process, researchers from the Massachusetts Institute of Technology (MIT) are honing a set of tools that would allow anyone to customize and create their own knitwear “without a memory bank of coding knowledge.”

One of them, a system called InverseKnit, leverages deep learning to translate a photo of a knitted product—a pair of fingerless gloves, for example—into instructions a machine can follow to output the design. Tests of InverseKnit have produced patterns with up to 94 percent accuracy, according to Alexandre Kaspar, a PhD student at MIT’s Computer Science and Artificial Intelligence Laboratory, and lead author of a new paper about the project.

“As far as machines and knitting go, this type of system could change accessibility for people looking to be the designers of their own items,” Kaspar said in a statement. “We want to let casual users get access to machines without needed programming expertise, so they can reap the benefits of customization by making use of machine learning for design and manufacturing.”

While InverseKnit currently operates on a small sample size, Kaspar and his team say they hope to employ the technique on a larger scale. They also plan to move beyond acrylic yarn by testing an assortment of materials to “make the system more flexible.”

Related Stories

A second project, known as CADKnit, uses a combination of 2-D images, computer-aided design software and photo-editing tools to allow users—non-expert ones, in particular—to tinker with patterns and shapes on knitwear, say by adding triangles to a beanie or vertical stripes to a sock.

“Whether it’s for the everyday user who wants to mimic a friend’s beanie hat, or a subset of the public who might benefit from using this tool in a manufacturing setting, we’re aiming to make the process more accessible for personal customization,” Kaspar said.

In surveys following tests, users said they found CADKnit’s templates easy to manipulate, though lace patterns proved tricky to simulate realistically. Whole-sweater customization, researchers added, are likewise a work in progress: While the design software is able to connect the trunk and sleeves in various ways, it doesn’t have a way to describe the “whole design space” as yet.

But these are early days, and the technology only has room to grow.

“The impact of 3D knitting has the potential to be even bigger than that of 3D printing,” Jim McCann, assistant professor in the Carnegie Mellon Robotics Institute, said in a statement. “Right now, design tools are holding the technology back, which is why this research is so important to the future.”

The demand for personalized knitwear certainly helps. Just last year, London’s Unmade raised $4 million in capital to bring its customization software to a larger audience, one that can feel “more involved and engaged in the experience and product” without shedding the speed and efficiency of mass manufacturing.

“There’s a clear and growing need for the Unmade business, at a time when trend life cycles are becoming ever more rapid, the desire for bespoke customer engagement is at an all-time high and brands are seeking diversified production, along with a strong, loyal community,” Frederic Court, founder and managing Partner of Felix Capital, said of the B2B startup, which has collaborated with companies such as Christopher Raeburn, Farfetch, Moniker and Opening Ceremony.