Live now in the eBay iOS app, you can tap the camera icon in the search bar at the top of the home screen, choose “Image Search,” then take a picture or upload one from Photos. You’ll have to grant eBay access to both your iOS Camera and Photos apps for the feature to work properly.
After you take a picture or choose one, you can crop it so it focuses on the item you want to search, and then tap “Search.” eBay will then show you listings for items that are a “close match” or “visually similar” to the product in your picture. Image Search takes into account all of eBay’s over one billion listings, according to the company.
Leveraging the latest advances in two core parts of artificial intelligence — computer vision and deep learning — these new features make it easier to find and buy the things that inspire you. When you upload images to run Find It On eBay and Image Search, we use a deep learning model called a convolutional neural network to process the images.
The output of the model gives us a representation of your image that we can use to compare to the images of the live listings on eBay. Then, we rank the items based on visual similarity and use our open-source Kubernetes platform to quickly bring these results to you, wherever you are in the world.
eBay also launched “Find It On eBay” for Android smartphones only, which lets users start an image search on another social platform, like Facebook. Using a share extension on an image, Android users can tap eBay and the retailer’s app will open to show similar listings.
We’ve performed a few quick Image Searches today and found mostly solid results for some items (an Apple Watch and band), although other products stumped the software so much we got an error (Nintendo Amiibo, Apple Watch dock). eBay still hasn’t said if some products are more reliable than others, but it did mention that Image Search will gain intelligence as customers use it more, as is typical with machine learning technology.
Image Search is live now on iOS and Android, and Find It On eBay has also launched on Android this week.