How Old does Your Computer Think You are?
Computerised face recognition is an important part of initiatives to develop security systems, in building social networks, in curating photographs, and many other applications. Systems that allow a computer to estimate with precision a person's age based on an analysis of their face are discussed in the International Journal of Applied Pattern Recognition.
Jayant Jagtap and Manesh Kokare of the Department of Electronics and Telecommunication Engineering, at the Shri Guru Gobind Singhji Institute of Engineering and Technology, in Vishnupuri, Nanded, India, suggest that age classification adds a useful layer to such security systems, customer relationship management, and of course for surveillance.
One of the best examples of face recognition becoming a widely used method for security come from China. China is developing what is expected to be the world’s most powerful facial recognition system with the capacity to identify any one of its 1.3 billion citizens within three seconds, according to a report by South China Morning Post.
Launched by the Ministry of Public Security in 2015 and developed in conjunction with a Shanghai-based security firm Isvision, the system can be connected to surveillance camera networks and will use cloud facilities to connect with data storage and processing centers distributed across the country, according to sources familiar with the project.
In line with what the researchers from India are saying, sources familiar with the Chinese project say there is no real indication as to when the system would be completed since the development was experiencing several issues related to the technical limitations of facial recognition technology and the large population base.
The team from Shri Guru Gobind Singhji Institute has carried out a detailed survey of age classification systems to reveal the pros and cons of each and to point new research in the right direction for the development of an even more accurate algorithm than any that currently exists. Indeed, their survey reveals that despite the best efforts of developers there is no real-time highly accurate algorithm for age classification yet in existence. The existing age classification systems commonly use geometric ratios of facial features and analysis of wrinkles in the skin.
One of the problems facing developers of such a tool is that databases containing images of a person's face tend not to accumulate images at different ages. This limits how well an algorithm might be trained based on a database of photos at known ages.
For the project in China, recent tests of the facial recognition algorithm, developed by Tsinghua University, saw an accuracy of below 60 percent. Meanwhile, the accuracy rate remained below 70 percent for the top 20 matches, as reported in a paper published in a Chinese journal called Electronic Science and Technology.
“Among 1.3 billion people, some totally unrelated people have faces so alike even their parents cannot tell them apart,” a researcher at the Institute of Computing Technology familiar with the project said.
While there are many serious tools and many spurious or fun tools available on the internet, there are no public systems that allow a computer to guess how old you are with useful accuracy. This may well be a positive point to make from the personal privacy point of view but not for the wider implications of face recognition and age categorization.
