GENERATIVE AI TECHNIQUES FOR ADAPTING STYLE OF A SPACE

Smart overview of the Invention

The patent application introduces a method utilizing generative AI to transform the style of a space through image generation. By leveraging a machine learning model, users can input an image of their space and specify a desired style, leading to the creation of new images reflecting that style. The system identifies furnishings in these generated images and matches them with similar products available in a catalog, providing users with detailed product information via an interface.

Background

Online retail platforms offer extensive product selections compared to traditional brick-and-mortar stores, providing consumers with millions of options. This vast array is accessible through web interfaces or dedicated applications, allowing customers to search for, purchase, and have products delivered directly to their homes. The application addresses the need for enhancing online shopping experiences by integrating advanced visualization techniques.

Generative Model Techniques

The system involves processing an original image of a space alongside a textual prompt that describes the target style. This prompt guides the generative ML model to produce one or more styled images. The model detects furnishings within these images and searches for similar items in a product catalog. This identification process involves comparing image portions containing detected furnishings with catalog images to find matching products.

Image Modification Process

Beyond generating styled images, the method also allows for modifying existing images by replacing specific furnishings with alternatives that fit the desired style. Users provide modification details, prompting the ML model to generate modified images. The system then identifies alternative furnishings in these images and matches them with products from the catalog, offering users relevant product information.

User Interface and Interaction

The application features user interfaces that facilitate interaction with the generative model. Users can view generated images alongside lists of similar furnishing products. Interfaces allow users to toggle between original and generated images or highlight specific furnishings for detailed views of matching products. Additionally, users can adjust prompts or seed values to refine image generation outcomes.