Classy “catalogue of ideas” app Pinterest has debuted an innovative new visual search system.
The Pinterest app has become a hub for community-sourced inspiration, with users pinning images of cool new items, furniture, and clothing.
Now a new feature makes identifying precisely what’s in those shared images even easier. When you spot something on Pinterest that you want to know more about, you can tap a search icon and draw a box around the part of the image you’re interested in.
It could be a lamp on a table, for example, or a pair of shoes being worn by a model.
Pinterest will then identify the object and supply any information it has associated with it, including the where to buy it if available. Even if there’s no associated information, Pinterest will show you pins featuring similar items. It’s also possible to filter your search to help narrow down the specific item.
In a related blog post, Pinterest engineer Andrew Zhai explained that the system was built in just a few months by a team of four engineers in close collaboration with the Berkeley Vision and Learning Center. It uses deep learning of Pinterest’s “richly annotated dataset of billions of Pins curated by Pinners” to identify specific objects.
This new visual search tool should now be available on the Pinterest mobile apps and the Pinterest website.