Technology is advancing at a fast pace. With the introduction of 3D printing machines, companies can easily create whatever needed to build the perfect product. While this technology and AI have been used extensively in most industries like automobile, marketing, etc, one area remains untouched; clothing. Being one of the 4 ‘essentials’, there’s no denying that clothing is a necessity. Which is why it was quite shocking when there were no businesses attempting to make this niche more efficient. However, not anymore. Researchers at MIT are working on an AI-based knitting software that will allow consumers to design their known knitwear. What’s even better is that even if you’ve never lifted a ball of yarn in your life, you can still effectively create items like gloves and sweaters, using this technology.
AI-Based Knitting: The Journey So Far
According to a paper released by a team of researchers at MIT’s Computer Science & Artificial Intelligence Laboratory (CSAIL), the team is building two software, InverseKnit & CADKnit. As quoted in the paper, computer scientist Alexandre Kaspar, the lead of this project says, “3D printing took a while before people were comfortable enough to think they could do something with it. It will be the same thing with what we do”.
The purpose of InverseKnit is to create patterns using photos of knitted items. To develop this system, the researchers first created a dataset of knitting patterns and matching images. This dataset was used to train a deep neural network, which was then able to generate machine knitting patterns. According to the researchers, the system has accurately followed the instructions provided 94% of the time. However, as the machine was only tested on one type of acrylic yarn, there is still work to do before the system can be commercialized.
Not too different from InverseKnit, CADKnit is software designed especially for people with little to no knitting or design experience. With this system, people can easily customize templates, adjust the size and final shape, all while adding decorative details as they like. CADKnit combines 2D images with photo-editing software in order to create customizable templates. In order to gain proof of concept, the researchers tested the system with people with little knitting experience. This test was a great success as, despite their limited knowledge, the candidates were able to create complex garments like gloves with lace motifs and different colour patterns.
While there is still more fine-tuning left, we believe that the MIT researchers are onto something revolutionary! Stay tuned on iTMunch to know more about the future of CADKnit & InverseKnit and for other tech updates