The statistics are brutal: 70% of users leave an online store if they haven’t found the product they need in a minute (Shopify). How can online retailers survive in such conditions and remain competitive? One possible solution is visual search. It makes the customer’s life easier and makes surfing the site much easier. Instead of text, there are now pictures, and store email data s have many new opportunities that will not only help them stay afloat, but also make good money.
Thus, visu ch is based on computer
Artificial intelligence has long been taking over the world. It has penetrated industry, medicine, education, computer games, and even creative fields such as marketing and journalism.
Ubiquitous neural networks have reached e-commerce. vision and deep
machine learning — a specially “trained” system of algorithms that can recognize products by images.
How was it taught? In a very exaggerated way, the process looks like this: the system is “fed” several million photos with images of a certain product. It analyzes its main attributes (color, size, material, silhouette, etc.) and remembers them.
Of course, this is not a matter of five minutes — it can take several months to train the system.
As a rule, neural networks undergo numerous tests to process data as closely as possible to the way the human brain would do it.
Search by color: similarity of the target-distractor; search by shape: set the size; search by color: set the size.
So far, such results have not been achieved, but everything is moving in that direction.
How it works
Instead of typing “red tight dress with a black zipper” in the input line, you just need to data on upload a photo/picture/screenshot of this dress to the website or application of the online store or paste a link to a post with it from Instagram.
The system first recognizes the image (reads its characteristic feat what new tld domains can we expect this year? ures and metadata), and then searches or similar images in its catalog, and, if necessary, directly on the Internet. And finally, it gives relevant possible products base
d on similarities – for example, color or style.
Ideally, visual search should find not only the desi duct, but also related products (similar or related products) and form a complete set. For example, to offer that same red dress, also shoes and a handbag.