Adobe Wants Consumers to Find the Perfect Shirt by Taking a Photo

New search tool links shopper photos to retailer inventory

Love the shirt, but don't know who made it? Adobe says it can help.
Getty Images

Software firm Adobe came to the NRF’s annual show earlier this week armed with a slew of product enhancements focused on what it calls customer experience management (CXM)—including some new visual search functionality for fashion retailers in particular.

This tool allows retailers to easily add functionality to their websites so customers can upload photos of clothes they like in order to find similar items within a given retailer’s catalog. It uses machine-learning tech Adobe Sensei in content management and data asset management platform Experience Manager, which searches inventory to deliver relevant matches.

Balaji Krishnamurthy, principal scientist at Adobe Experience Cloud, said Adobe is piloting the tech with a “major retailer.”

As a result, it has indexed said retailer’s product catalog using a deep-learning system that extracts relevant information. He said this, in turn, means it’s able to isolate apparel—even in cluttered images—which it demarcates with dots. When consumers see something they like, they can click on the dots to see relevant matches of items available for sale.

“It is sorted in order of how similar it is and also conveniently categorized into different apparel categories,” Krishnamurthy said.

Jonas Dahl, product manager at Experience Manager, noted this is a great way for retailers help consumers connect their own images to actual inventory. Examples include untagged apparel on social media, which might be hard to find otherwise.

It also means if a customer is looking for, say, a red dress with a floral print, a retailer can offer similar items even if it doesn’t have an exact match.

“We don’t want to go back and say, ‘No, we don’t have it, find it somewhere else,’ Dahl said. “We want to say, ‘We don’t have that exact thing, but why don’t you look at the other ones that are similar?’”

It also encourages consumers to browse further between categories.

“It’s a very engaging way of exploring the product catalog based on visually similar products, but it also turns out it’s good for recommendations,” Dahl said. “A lot of recommendation systems rely on ‘other people who viewed this also viewed this or bought this,’ which is great, but it turns out visually similar items can create a different kind of recommendation that you can blend into normal recommendations.”

Adobe will focus on fashion to start and will offer the feature to retail customers through an API.

Recommended articles