Tech to invade Wardrobes; Fashion industry gets AI-designs

Two things that are always evolving and are forever on the wish list of the millennial are Fashion and Technology. And two women-graduates from the Massachusetts Institute of Technology (MIT, USA) are combining both fashion and technology in a uniquely developed offering.

Pinar Yanardag and Emily Salvador have launched their fashion brand Coven.ai. Like any fashion brand, Coven.ai will sell apparels but its designs will not come from the Tokyo or Milan collection, or any other budding designer – but from an AI-powered machine.

This is not the first time when Artificial Intelligence has found its way into the fashion industry. Many fashion brands are using AI to predict the upcoming trends and understand the collective behavior.

But the duo at the MIT is re-imagining the dress designs with the help of an emerging artificial intelligence technique called the Generative Adversarial Networks (GANs).

GAN in dress-designing

Designers start with a general sketch on paper; then add styles, elements and colors; revise (and even refine) the designs; and finally handover their designs to the dressmakers.

GAN is made to understand that the pattern drafting is the primary and most important step in dressmaking. The GAN-based algorithm makes Coven.ai accelerate this time-consuming and labour-intensive process.

The company’s AI algorithms have produced two designs so far. It has designed a shimmering party dress and a formal black dress with sharp shoulders, asymmetrical neck and bell sleeves.

How Coven works?

The team at Coven.ai has designed nearly 30,000 high-resolution dress patterns, after downloading styles from 10 websites.

The Coven.ai team trained an AI model to randomly generate thousands of sewing patterns; then selected a few designs based on their tastes. It finally assembled fabrics and other materials to deliver the design.

It uses the process is StyleGAN (Style-based Generative Adversarial Network). StyleGAN represents an alternative generator architecture that draws insights from style transfer techniques. The algorithm may learn and separate different aspects of an image unsupervised. It enables intuitive, scale-specific control of the synthesis. StyleGAN is trained so it can learn how to synthesize new dress designs.

Inspiration

Yanardag initiated the project How To Generate Almost Anything  at MIT Media Lab. The project used AI algorithms to create pizza recipes, songs, dresses, perfumes, screenplays, graffiti, and even viruses.

Yanardag and Salvador were excited about its potential and wanted to try the process on dresses, believing AI-powered fashion could have potential.

Coven.ai has produced 100 affordable AI-designed dresses, and are planning to launch a kickstarter campaign for fundraising. The company wants to enable the creation of personalized dresses based on individual customer taste.

The duo are now working on adopting StyleGAN to add personalization directly as a feedback loop to the algorithm.

Can AI endanger job of designers?

The duo says that Coven.ai does not intend to replace fashion designers. It is only considering building an AI tool to reduce fashion designers’ workloads. It may offer automatic preview options for a wide range of patterns and styles in a particular design.

The Coven.ai team has also collected a number of sketches from internet archive of the Metropolitan Museum of Art. Thereafter, it regenerated the dress designs using style transfer, an AI technique for recomposing images in the style of other images.

Since Yanardag and Salvador are from the non-fashion backgrounds, Yanardag is now attending fashion school on weekends to build effective manufacturing, retailing, and marketing. She is also hunting talents with fashion business experience at the school who may join Coven.ai’s team