In cooperation with our friends at ETH Zurich we have been working on automatic image recognition for places such as landmark buildings for a while. We think it reached a level, where it can be of benefit to our Déjà Vu users. Today, we release recognition of landmarks as an Alpha feature in Déjà Vu.

How it works:

• All your images will be looped through our new landmark recognition service
• The recognition service currently covers a few hundred thousand landmark buildings, located mostly in Europe, the US, and Japan
• If a landmark is recognized in Déjà Vu it will be added as a match. If you choose the match, it will add a title, categorize the image as a “Place”, and will even place related Wikipedia links for the place
• The main challenge here is reference database size, i.e. coverage. If we don’t recognize a building it’s in most case due to lack of data. We are working on expanding the data coverage with full power. But it will take a few months, as we have to crawl basically the whole world for relevant landmark images.

The images below show a few examples of how this looks in action:

Why is this useful for you? Well, do you remember your last holiday trip? Did you take pictures of places or sights? Most certainly! But if you browse your collection today, of how many places are you still able to tell their name? (Of course, besides the obvious cases like the Eiffel Tower). You will notice that it’s actually quite hard. And here Déjà Vu intends to help you out.

Try it yourself:
You could start with the sample images below, or use your own. (By the way, of how many of the images below you know what they are? One for sure – maybe two.)