Modern AI System To Reveal Gender & Age Of People On Videos

An artificial intelligence (AI) system has been devised by scientists of Russia’s Higher School of Economics to detect gender and age of people on video with the high accuracy. This technology is expected to apply into offline detection systems of the apps which run on the Android operating system.

Advanced neural networks integrated on the system analyze and find out the gender of members on videos. Its accuracy can reach up to 90%. However, it’s much more difficult to detect the precise age.

For traditional neural networks, each age value on each video frame will be considered to calculate the most possibly exact age. Let’s see the following example: when analyzing a video, the traditional system identifies that a person is under 21 years old in 30% of the frames, and under 60 years old in 10%. Comparing between these values, it will give a conclusion: this person is estimated 60.

However, the result of the age prediction depends much on a series of different conditions, like the head rotation or how to observe each video frame. That’s why we should add or reduce about 5 years when calculating the age of the same person through various frames.

Everything has changed with the modern technology developed by scientists from the Higher School of Economics in Russia. They have introduced a new solution to enhance the performance of neural networks by combining confidence levels which the neural networks offer in each video frame.

As we know, the software systems which can analyze facial recognition usually contain discrete neural networks. Among these networks, one can detect the gender while the other can identify the person.

The modern neural network offering a lot of outputs is now highly appreciated. Some features of the multi-functional system include prediction of gender and age, production of a 1000-number set for each person’s unique attribution, as well as distinguishing a person from others.