AI Visual Search for Home Décor

Recently, a number of specialized search engines focused on visual and stylistic aspects of fashion have emerged, designed to assist consumers in locating desired items for online purchase. These platforms utilize computer vision and other artificial intelligence technologies to enable more effective visual searches, accurately identifying and presenting specific designs and styles.

London-based Cadeera, currently in the process of securing seed funding, is developing a similar AI-powered visual search solution, but with a focus on interior design. Throughout the pandemic, the company has been refining a prototype aimed at enhancing the discovery of aesthetically-driven products such as sofas, armchairs, and coffee tables within the e-commerce landscape.

Sebastian Spiegler, the Founder and CEO, who previously worked at SwiftKey and holds a PhD in machine learning and natural language processing, demonstrated the prototype to TechCrunch.

The software features a user experience designed to initially determine a user’s stylistic preferences for interior décor. This is achieved by presenting a series of room images in various décor styles and asking the user to indicate their approval or disapproval using a swipe-based interface similar to that of Tinder.

Based on these preference signals, the system begins to suggest specific items for purchase – for example, armchairs or sofas – that align with the styles the user has indicated they like. Users can further refine the selections by requesting similar items (“more like this”) or broadening the range by requesting less similar items, introducing an element of chance into the search process.

The platform also allows users to initiate searches by uploading an image, with Cadeera then identifying and displaying comparable items currently available for sale.

An augmented reality feature is also planned for future implementation, enabling users to visualize potential purchases within their own homes. Voice search functionality will also be incorporated.

Spiegler contends that “keyword search is fundamentally flawed.” He explains that when renovating or redecorating, individuals often have a vague idea of what they want, knowing it when they see it but lacking the specific terminology to describe it. The challenge, therefore, is to help users articulate their desires.

“Our core objective is to understand individual preferences. If someone is unsure of what they are looking for, we aim to provide visual cues, personalization, and inspirational content – including images and, eventually, community features – to help them define their needs. This also benefits retailers by providing insights into customer preferences.”

“This approach fosters greater confidence in purchasing decisions, reduces the likelihood of returns – a significant expense for retailers – and may even encourage increased spending as users more easily discover items they desire,” he adds.

The pandemic has significantly accelerated the growth of e-commerce, driving more shopping activity online. However, this has simultaneously presented challenges for traditional brick-and-mortar retailers.

This situation is particularly acute for furniture retailers who previously relied on physical showrooms to facilitate product discovery and sales. Consequently, they are actively seeking innovative tools to enhance their online presence and boost sales.

The emergence of sector-specific visual search engines appears poised to benefit from this broader shift towards e-commerce driven by the pandemic.

“I believe the interior design and home décor market is currently underserved,” Spiegler stated to TechCrunch. “In my opinion, no existing solution has effectively addressed the challenge of searching and discovering items easily. While several companies operate in the fashion sector, I am confident we can excel in the furniture and home décor space and then expand into other areas. The opportunity is substantial.”

“We can adapt many of the successful strategies from the fashion industry to the furniture market,” he continues. “I believe there is a significant gap in the market that has not been adequately addressed.”

Spiegler estimates the global market opportunity for Cadeera to be between $10 and $20 billion.

The startup’s initial business model is focused on B2B, with plans to offer its Software-as-a-Service (SaaS) to e-commerce retailers, enabling them to integrate the visual search tools directly into their websites.

Spiegler reports that they are collaborating with a “major” U.K.-based vintage platform and aim to launch a product to market within the next six to nine months, initially with one or two customers.

They also intend to develop applications for popular e-commerce platforms such as WooCommerce, BigCommerce, and Shopify, providing a suite of their search tools. (Larger retailers will have access to more customized platform options.)

Regarding the possibility of developing a direct-to-consumer application, Spiegler acknowledges that it is a “long-term goal”.

“Ultimately, my vision is to create a consumer app – a central hub for all your shopping preferences. It would be a blend of Instagram’s inspirational content and Pinterest’s organizational features, providing relevant recommendations,” he says.

“This is the concept behind the product search engine we aspire to build. However, the current steps involve a B2B approach, which we believe will eventually lead to a community-driven B2C app. We aim to begin with B2B and then transition to a consumer-focused model.”

For the time being, the B2B strategy allows Cadeera to collaborate closely with retailers, gaining valuable insights into market dynamics and accessing essential data to power its platform, such as style lookbooks and item databases.

“This will result in a comprehensive inventory database, a design knowledge base, and a collection of images and style metadata. We will then apply AI technologies – including object detection, object recognition, recommendation systems – to facilitate personalization, exploration, search, and suggestions,” he elaborates, outlining the various technical components.

“We also provide an API for seamless integration and, if needed, a responsive UX/UI.”

“Furthermore, we are building a valuable data asset that understands user preferences, creating user profiles that can be leveraged across platforms. This means that a purchase made on one e-commerce site can inform recommendations on another, providing users with the most relevant suggestions based on their purchasing history, style preferences, and interaction data,” he adds. “Your complete purchasing history, style preferences, and interaction data will allow you to get the most relevant recommendations.”

While established tech companies like Google and Amazon dominate the broader search and e-commerce markets, Cadeera is not concerned about competition from these global giants, given their lack of focus on the specific needs of the furniture and home décor niche.

He also notes that Amazon consistently struggles with providing effective recommendations, despite possessing a vast amount of data.

“I have repeatedly asked – and been asked myself – why Amazon’s recommendations and search results are so poor. The honest answer is I don’t know! They likely have the best data set available… but the recommendations are subpar,” he says. “Our goal is to completely reimagine the product search experience. Search should function effectively, and the inspirational aspect, particularly for items that are difficult to define, is a crucial element missing from everything I’ve seen so far.”

And while Facebook acquired a home décor-focused visual search service (GrokStyle) in 2019, Spiegler suggests that Facebook is likely to integrate that technology – including AR visualization – into its own marketplace, whereas he believes most retailers will prefer to maintain their independence from the Facebook ecosystem.

“GrokStyle will become part of Facebook marketplace, but for retailers, the key question is how much integration with Facebook is desirable, and how much dependence on Facebook is acceptable? I believe that’s a significant concern for many retailers. Do you want to rely on Google? Amazon? Facebook?” he says. “My prediction is no. Because they ultimately risk being overshadowed.”