An AI (artificial intelligence) system has been developed by researchers that can recognize individuals on video, identifying their gender and age more accurately and quickly. Already, the development has turned out to be the foundation for offline recognition techniques in Android mobile applications, as per the scientists from the Higher School of Economics, Russia.
Advanced neural networks identify gender with an accuracy of 90% on videos. The condition with age calculation is much more complex. Traditional neural networks mull over distinct age values. In every video frames, the system calculates the odds of the individual in the photo being of a particular age.
Owing to several circumstances of examinations or even trivial head movement, calculation of the age of that identical individuals in diverse video frames differs in the array of 5 Years, minus or plus. Scientists discovered a means to enhance the operations of neural networks. They applied a novel technique to combine confidence levels generated by the neural network for every frame.
Usually, the software systems of facial identification analysis comprise numerous distinct neural networks. Among them, one recognizes the individual, while another identifies the gender, and so on. An efficient neural network with numerous results has been developed.
It resolves numerous tasks simultaneously: identifies gender & age and generates a set of 1,000 numbers that exclusively assign every individual and enables them to be differentiated from other individuals. As per the scientists, this solution will function even on low-performance smartphones.
Likewise, IBM scientists have designed a first-of-a-its-type “fingernail sensor” model that utilizes machine learning and AI to supervise and examine human health and disease progression. The wireless, wearable device constantly gauges how an individual’s fingernail moves and bends, which is a major marker of grip strength.
This new system utilizes signs from the fingernail bends such as the perceptible sensing of pressure, surface textures, temperature.