• Philipp Derksen

Images search better than words


As an image search technology Visual Search has revolutionized online commerce and solved some major problems for users and retailers alike. Product recommendations and search based on images make the shopping experience more intuitive and emotional for users while allowing retailers a more effective and lucrative business.

In this series, we will present different applications for Visual Search in e-commerce. Images search better than words

Searching with images has become part of everyday life. The smartphone camera has become an input device and navigation aid. With Alexa, Siri and other voice assistants turning the microphone into a much-used input tool, the use of the camera with Lens and Co. is now increasingly becoming a learned cultural technique.


This is how the AI-based image search works: Users upload a product photo - a self-made picture of a handbag or an Instagram screenshot of a sofa - by clicking on the camera icon of an online shop, receiving suggestions of stylistically identical or very similar products as a result.

These hits are more precise than any text- or behavior-based search. A picture is worth a thousand words. And it makes for a better search than any number of words you might use.

Products that cannot be named or described exactly can be found more quickly with visual search. Often, the conventional search is hampered by imprecise terms and missing synonyms: Is the desired handbag a handle bag with a rounded top or rather a half-moon shaped hobo bag? Is the sofa you're looking for a two-seater? And how do you describe those special decorative cushions?

Visual search, on the other hand, is intuitive and fast. It substantially simplifies the search entry, takes some burden off customer service and is, in essence, a 24/7 mobile booster for any online business.

Our software vviinn takes this intuitive product search to another level. We have developed a special image analysis process that allows vviinn to analyze product images based on more than 1,000 features and suggest stylistically suitable products within 100 milliseconds.