06 December 2015, 23:36 / Technologies

Russian startup beats Google team

Together they participated in a facial recognition algorithm competition Elizaveta Sergina

Modern algorithms make it possible to identify more than 70% of individuals / Michaela Rehle / REUTERS

The University of Washington summed up the results of the Megaface facial recognition algorithm competition (published on the website; a university representative did not comment on them). N-tech.lab entered the top five in all four categories and won in two, beating the Google team as well.

Participants in the competition were asked to “recognize” images of people from two collections in a large database of photographs. The first collection consisted of 100,000 images of people of approximately the same age, the second - of 900 people of different ages. The search was carried out in a database of 1 million images.

Intel Corporation announced a new, improved processor on Tuesday.

Technology / IT business

There were four nominations in total: each collection had to be assembled from two databases – a larger and a smaller one. 100 teams registered in the competition, each team could participate in two categories: “heavy” - without restrictions on the number of photos for training the algorithm, and “lightweight” - with a limit of 500,000 photos.

The N-tech.lab algorithm from the “heavyweight” category was able to recognize more than 73% of the proposed faces from the first collection in a database with 1 million photos (Google - just over 70%), follows from the results of the competition. In the “lightweight” category, N-tech was second. In recognizing 900 images from the second collection in the “heavy” category, Google turned out to be the best (identification level - 74%), N-tech.lab - second (52%). N-tech beat its rivals in searching 900 images in the “lightweight” category.

“We consider this a victory, because our algorithm showed the best result in the main category,” says N-tech.lab founder Artem Kukharenko. “It’s especially nice that we trained our algorithm on three computers, while Google needed 1,000 servers.”

Typhoon from the journal

The website of the Typhoon digital development fund says that it was created at the end of 2013 as a result of the consolidation of several Internet companies, manages a portfolio of assets with an estimated value of more than $10 million. All three partners are graduates of the Faculty of Journalism of Moscow State University: Bratishko worked in the newspaper Vremya Novostei, was the editor-in-chief of the publication “Our Version”, Gusev worked in the editorial office of NTV news programs, the newspapers “Izvestia” and “Novye Izvestia”, and participated in the creation of the magazine “Moscow: instructions for use”. Kabakov founded the information and analytical portal RIA “Vybory”. In 2010, they founded the PR company Agency One.

N-tech.lab was founded this year, according to its website. Kukharenko is a graduate of the Faculty of Computational Mathematics and Cybernetics of Moscow State University, but the share of the founder of the company, according to SPARK, is 25%. Another 17.5% each belongs to the managing partners of the Typhoon digital development fund Oleg Bratishko, Anatoly Gusev, Alexander Kabakov (see inset), as well as Marina Polyanskaya, 5% to Ekaterina Semenova.

Prizes in the competition will not bring money to the company, but Kukharenko is counting on something else: the Computer Vision and Pattern Recognition Conference will be held in June 2016, at which the University of Washington will make a report on the achievements of facial recognition technology, including the results of Megaface. This will attract the attention of N-tech investors in whom the company is interested, Kukharenko does not hide.

Facebook is creating a virtual assistant

Technologies / Internet and digital

According to him, the company is preparing to soon launch several commercial products using its algorithm. One of these projects is in the entertainment sector. This is a dating service that allows you to search for people with a certain type of appearance.

Facial recognition algorithms are an extremely promising industry, Kukharenko is sure, because they can help catch criminals in a crowd (if there are images from surveillance cameras - for example, in the subway), this algorithm can also find application in retail (customer recognition).

The directors of Typhoon learned about Kukharenko from mutual friends, Kabakov said: “We looked at the project, it immediately became clear to us that the team had unique experience and a product with good prospects in the field of entertainment and information security, and decided to invest.” According to the investor, the project is not specifically looking for anyone, but if there are people willing to invest in it, the shareholders will consider this proposal.

Yandex bought a startup in Israel

Technologies

For startups, participation in such competitions is very useful: there are investors who are guided by the results of such competitions, says Alexey Soloviev, managing director of the Prostor Capital fund.

In matters of seed investment, the availability of a basic product, the quality of the startup team, as well as the potential of the target markets are of fundamental importance, said Sergei Skripnikov, a representative of the Internet Initiatives Development Fund (IIDF). As a seed fund, IIDF works with Typhoon, but has not yet heard of N-tech, he says.