How Does Search by Photo Work for Clothing?
Search by photo analyses the category, colour, pattern, neckline, sleeve and fabric look of an uploaded garment with AI and matches it against similar items in the catalog. MerterAI combines vector similarity with AI tags; the result list leads all the way to supplier comparison.
What happens technically?
The uploaded image is converted into a numeric vector. Every catalog product has a vector in the same form. Search finds the nearest neighbours between these vectors. Results are then filtered and ranked with AI tags (category, colour, pattern, style).
How is it different from text search?
Text search matches the user's phrase against product tags and is fast for describable features such as 'polka dot dress'. Visual search works on patterns whose names are unknown: finding the equivalent of a dress seen on social media requires no description at all.
How is it used in wholesale buying?
Boutique owners search images from competitor windows, social posts or customer requests to find the manufacturer of the same model. The similar-products list on the matching page shows how to source the same model from different suppliers.
| Feature | Visual search | Text search |
|---|---|---|
| Input | Product photo | Natural-language phrase |
| Match basis | Silhouette, colour, pattern similarity | Tag and title matching |
| Strong at | Styles with unknown names | Describable features |
| Combines with | Category/colour/pattern filters | Category/colour/pattern filters |
| Output | Similar products + supplier comparison | Similar products + supplier comparison |
Frequently asked questions
Which image formats are supported?
Common image formats (JPEG, PNG, WebP) are supported. Clear single-item shots give the best results.
How can results be improved?
Adding category and colour filters, using a detail shot and narrowing by silhouette raise match quality.
Does visual search work on mobile?
Yes, the platform runs in mobile browsers; images are uploaded from the camera or gallery.