March 26, 2021

Top Face Recognition Research in the World

Biometrics Research Laboratories
Takaya Miyamoto

After receiving a doctorate in physics, Miyamoto joined NEC in April 2019. During his student days, he majored in theoretical particle physics but has been engaged in face recognition research since joining the company. He received the MIRU Interactive Poster Award at the 23rd Meeting on Image Recognition and Understanding (MIRU2020).

Increasing the value and uses of face recognition

I currently research face recognition technologies at NEC's Biometrics Research Laboratories. NEC is a company with some of the world's best technologies in biometric authentication. In particular, the company has been ranked number one in the world* five times in face recognition benchmark tests hosted by the U.S. National Institute of Standards and Technology (NIST).
However, in recent years the deep learning wave of technology innovation has spread throughout the world, and the competition in the development of face recognition systems is intensifying. Of course, NEC is also not currently resting on its laurels. While scanning new research papers on a daily basis and striving to research and develop new technologies, the team is united as one in continuing its research.
I am mainly in charge of basic research to increase the utility value and added value of face recognition. While technology research to increase the fundamental accuracy and speed of face recognition is certainly important, because there is generally a trade-off relationship between accuracy and speed, optimizing the accuracy and speed according to the operation scenario becomes important. For example, these days even smartphones are equipped with face recognition systems, but high-speed calculations cannot be carried out on such small edge devices unlike on servers, and the recognition speed decreases. Instead of providing recognition with high accuracy, it can take several seconds, and the users feel stressed. Well I just used smartphones as an example to explain the issues in an easy to understand manner, the construction of a system that can complete the recognition within a small edge device has a wide range of many other usage applications such as company entrances and outdoor gates that are not connected to a network.
Moreover, thermo cameras for infection control measures and infrared cameras for nighttime crime prevention are recently being used in an increasing number of locations. The development of a face recognition system which provides high accuracy and high speed even with images that would be difficult to judge with the human eye is likely to have significant value. My goal is to further increase the utility value and brand value of NEC's face recognition technologies by developing face recognition that can adapt to a wide variety of scenarios.

Assembling more than 100 exceptionally talented people

NEC's Biometrics Research Laboratories currently has over 100 researchers. Among laboratories focusing on biometric authentication research, there are probably not very many organizations in the world with such a large scale.
Each individual researcher is an exceptional person with unparalleled expertise. The company has assembled people who are well-versed in algorithms, hardware experts, those who are unmatched in programming and software implementation, and other specialists with knowledge and know-how that is the best in the world in various areas. It is a unique place in that many people playing an active role hold a doctorate. There are many who joined from different fields such as one researcher who joined after producing results in the field of astrophysics and another who joined after completing a doctorate in biology at an overseas university. Another interesting aspect is that there are researchers who come not only from the sciences but also from the humanities.
The 100 person strong group of specialists with such a wide variety of backgrounds is conducting research from their own respective approaches and engaging in mutual debates on a daily basis while researching. Taken as a whole, the group brings together an amazing amount of brainpower. I believe that such a research environment is what supports NEC's biometric authentication technologies, which are among the best in the world.

Research which anticipates technology applications

When I was in the final stages of my doctorate, I was faced with the two paths of either continuing to remain in academia or to become a corporate researcher, and I was extremely conflicted. In order to reach a conclusion, I postponed my graduation for almost one year. I was that conflicted about the situation.
The answer that I finally reached as a result was that I wanted to conduct research in a form that would be deployed in manufacturing and society. The technologies that I made might change the lives of everyone around the world and make them happy. I felt that would be the happiest outcome. I think that the true appeal of corporate research is the ability to experience the satisfaction of having your technologies become products and services that spread throughout society.
However, a different perspective from academic research is required in order to do so. A research posture which anticipates not only the technology novelty and scientific interest but also how the technology will be useful to society becomes important. The most desirable outcome is to predict the final application, but depicting such a far-reaching vision from the research stage is quite difficult.
For that very reason, what I now view as important is the ability to explain things in a way that even regular people and non-specialists will think, "that's amazing!" If you can explain the value in an easy to understand manner, people from the Business Division will also understand the significance of the technology and point out the needs of front line users and useful ways to deploy the technology. The mutual discussion will also proceed in a smooth and constructive manner.
The ideal outcome is to be able to take a technology that gets written up in a research paper accepted at a prestigious international conference and further productize and commercialize it. Going forward, I would naturally like to continue to conduct cutting-edge research as well as advance technology research that may significantly change society through collaboration with the various divisions and a research posture which anticipates commercialization.

A day at work

Message to my past self in my school days