3D printing has made countless items readily available, but there’s usually a catch. For most novice technology users, open-source CAD drawings are the only way to access product designs, as customizations — even minor modifications — are unattainable. That’s because the meticulous nature of additive manufacturing means any design updates would require a new, modified CAD drawing. Even if the user were capable of manipulating the design, it could risk the functionality of the product.
With this challenge in mind, MIT researchers have been exploring ways to apply artificial intelligence (AI) to the practice of modifying 3D models. A team of researchers is experimenting with a tool they call Style2Fab, designed to enable users to personalize items without compromising their functionality.
According to Faraz Faruqi, author of a recent paper on Style2Fab and a graduate student in computer science at MIT, Style2Fab “would make it very easy to stylize and print a 3D model, but also experiment and learn while doing it.”
The technology functions by utilizing a deep-learning algorithm that can segment the design, identifying which parts are functional and which are purely aesthetic. It explores the model’s topology and compares its segments to a database of other functional 3D models. The parts that match stay. A human actor is also involved in assessing the comparison, at which point the user can apply “Text2Mesh,” a second system that applies AI to the aesthetics.
The team believes that medical customization could be a huge market for this technology because a patient who can customize a device is much more likely to use it. The report cites the example of a thumb split that can be adapted to match the wearer’s clothing — a steep challenge for a clinician to produce but a potentially manageable fix for Style2Fab.
More AI Insights
Image Credit: MIT
