Searchface: Russian developers claim that their new face search engine is more effective than others

Searchface: Russian developers claim that their new face search engine is more effective than others

Several days ago Russian developers unveiled a demo of a new face search engine that allows searching Vkontakte users by photo. Vkontakte is the largest Russian social network with over 97M active users. Earlier the site would give a link to user's profile but after Vkontakte has filed a suit against Searchface, the link was removed.

In an article on (machine-traslated) developers claim that they had improved the algorithm so that the face can be recognised even if it is partially covered or has low quality.

The engine can find the photo even if the person later has restricted access to their profile; the developers have collected the database of photos without asking for anyone's permission.

Here are examples of search results with a partially covered face:

Maxim cherkasov,

Despite face covered with sunglasses and despite that some of the photos had low resolution, the seacrh engine has produced correct results:

Ilya Krasilshik from has tried searching with a photo taken from the side. The algorith was able to find matching photos:

Here is another example for seacrh by partially covered face:

Developers also claim that their algorithm can find a face if it is blurred:

Here are some more examples:

Developers also claim they can find person by a photo from their childhood. Here are the search results for 8 years old photo of Amy Winehouse:

While algorithm had difficulties with 7 years old photo of Madonna, it could still produce some results:

Developers ask everyone to try out their search engine themselves: users also verified that the engine was able to find Tom Cruise with a makeup:

This is not the first face search engine; another publicly available search engine was findface, but later it had closed public access and the company moved to developing private solutions for face recognition. А Russian photographer had successfully used it to identify strangers in the subway to prove its efficiency; a new search engine claims to yield even better results.

Another use for findface was to identify porn and webcam models; the users of Russian imageboard 2ch found women' accounts in the social network and told about their occupation to their friends, relatives and the administration of the college or the university where they were studying at.

Findface was also used by a pro-government group to identify participants of a protest demonstration and publish their data on the website "jesuismaidan".

Both findface and searchface collect photos from social networks without getting permission from its users; findface required the user to provide access to their VK account and it allowed findface to collect photos closed from public view. Some time ago Twitter closed access to its API for findface; but this won't prevent it from collecting publicly accessible photos.

There is no doubt that in future there will be more demand for such kind of applications: police will want to identify criminals and protesters, stores will want to identify their customers. For example, NVIDIA's article says that face recognition products will find its use at "malls, concert venues, casinos, or any place where many people congregate and there are advantages to being able to track them individually".

Whether the consumers will be able to protect their rights still remains an open question.