In the ever-evolving realm of artificial intelligence, groundbreaking technologies are continually pushing the boundaries of what’s possible. One such innovative application that’s making waves in the fashion industry is the use of diffusion models to generate try-on images. This cutting-edge approach is revolutionizing the way we visualize and experience fashion across a diverse spectrum of individuals and poses.

Understanding the Diffusion Model:

At the heart of this technological marvel lies the diffusion model, a sophisticated technique employed by AI systems to seamlessly showcase clothing on a diverse range of body types and poses. So, what exactly is the diffusion model?

In the context of AI and fashion, diffusion refers to the process of simulating how clothing interacts with different body shapes and movements. It essentially allows the AI model to ‘diffuse’ or spread the visual representation of clothing across various individuals, capturing the nuances of fit, texture, and style.

How the Diffusion Model Works:

1. Representation Learning: The AI model undergoes training to learn a robust representation of clothing items. This involves analyzing vast datasets of images featuring various garments on different individuals, capturing the intricate details of fabric drape, creases, and style variations.
2. Pose and Body Variation: The model is then exposed to a diverse array of poses and body types. This step is crucial for the diffusion model to understand how clothing adapts to different postures and sizes realistically.
3. Simulation and Adaptation: Leveraging the learned representations, the diffusion model simulates the try-on process virtually. It adapts the clothing item to the specifics of a given pose and body shape, producing a remarkably realistic depiction of how the garment would look on a real person in that particular stance.
4. Generating Try-On Images: The final output is a series of try-on images that showcase the clothing in a way that transcends traditional static imagery. From dynamic action poses to relaxed standing positions, the diffusion model provides a comprehensive view of how the apparel complements the wearer’s body and movement.

Implications for the Fashion Industry:

The integration of diffusion models into the fashion landscape holds immense potential for various stakeholders:

1. Enhanced Online Shopping Experience: Consumers can make more informed purchasing decisions as they virtually try on garments, gaining a realistic preview of how the clothing will look on them.
2. Diversity and Inclusivity: The diffusion model celebrates diversity by showcasing fashion on a wide range of body types, contributing to a more inclusive representation of beauty standards in the industry.
3. Reduced Return Rates: By offering a more accurate representation of how clothes fit, retailers can potentially reduce return rates, leading to more sustainable and efficient operations.

Like Google, other tech companies and retailers have caught on to consumers’ desire for more reliable online shopping experiences. In recent years, Snap has used augmented reality for various virtual try-on efforts, including partnerships with brands like American Eagle, Gucci and Men’s Wearhouse. According to a Snap survey from June 2022, eight in 10 respondents are more confident in their buying decisions after using AR shopping tools.

Google introduces generative AI virtual try-on tool:

Google released its virtual try-on tool in response to online shopping frustrations. According to a survey Google conducted with Ipsos in April with 1,614 U.S. adults, 59% of respondents are unsatisfied with products they buy online because they look different on them than anticipated.

In its blog post, Google hinted at its plans to adopt AI technology to enhance online shopping for other items as well. The company said it aims to launch a virtual try-on tool for men’s apparel, including tops, sometime later this year.

“Our new generative AI model can take just one clothing image and accurately reflect how it would drape, fold, cling, stretch and form wrinkles and shadows on a diverse set of real models in various poses,” Lilian Rincon, senior director of product and shopping, wrote in a company blog post.

As shown in the above screenshot, the diffusion model sends images to their own neural network (a U-net) to generate the output: a photorealistic image of the person wearing the garment.

you can use virtual try-ons for apparel on women’s tops from brands across Google’s Shopping Graph, including Anthropologie, LOFT, H&M, and Everlane. Over time, the tool will get even more precise and expand to more of your favorite brands.

By Asif Raza