In a groundbreaking innovation for online shopping, Google has announced its latest feature – a virtual try-on (VTO) for apparel. This cutting-edge feature allows customers to gain a detailed view of clothing on models of different body sizes and shapes. Users can observe crucial aspects of the apparel such as draping, folds, stretching, clinging, and wrinkles using this feature. This is an unprecedented revolution in the online retail sector, brought to life by the relentless efforts of Google’s shopping AI researchers who have devised a new generative AI model.
Previously, VTO techniques such as geometric warping were used to adapt clothing images to a silhouette. However, these techniques often did not result in a natural-looking image. Google’s focus, therefore, has been on creating each pixel of a garment from scratch in order to produce high-quality, realistic images. This was accomplished through a novel diffusion-based AI model.
Also Read: Coca-Cola Appoints Pratik Thakar as Global Head of Generative AI in Pursuit of Marketing Transformation
Diffusion, in this context, is the process of gradually adding extra pixels to an image until it is rendered unrecognisable, and then subsequently removing this added noise until the original image is perfectly reconstructed. Text-to-image models like Imagen employ diffusion in conjunction with text from a large language model (LLM) to create images based solely on text.
But Google’s new approach to VTO deviates slightly. The diffusion process uses a pair of images as input—one of the garment and another of a person. Each image is processed through its own neural network and shares information through a process termed ‘cross-attention’. The final output is a photorealistic image of the person wearing the garment.
The AI model was trained using Google’s Shopping Graph, which is home to the world’s most comprehensive data set of the latest products, sellers, brands, reviews, and inventory. This model was trained using countless pairs of images, each showcasing a person wearing a garment in two different poses.
The procedure was then replicated using millions of random image pairs featuring various garments and people. This resulted in a tool that enables users to realistically visualise how a top would look on a model of their choice.
The VTO feature is currently available for women’s tops from brands including Anthropologie, LOFT, H&M, and Everlane across Google’s Shopping Graph. As the technology advances, the tool is set to become even more precise and will be extended to additional brands.
This news is based on the fibre2fashion website.
