Emotions are what makes our behavior unique. We can not ignore emotions if we want to build a trustworthy relationship between artificial intelligence and a human being.
The correct understanding of emotional background is the key to make the next step in communication. What is especially important in case of artificial intelligence is that it can keep attention on manifestations of emotions and measure mimicry.
These measurements are based on the movement of the landmarks on a face and can predict an outcome of the person’s mood.
Our approach permitted us to solve tasks of facial and landmarks detection with a minimum of computational resources and made our solution accessible even for mobile devices and efficiently processes not only images, but also a video. At the same time we provide 21 landmarks that allow a precise measurement of the mimicry. With our algorithm, we leveraged recent breakthroughs in training of deep neural networks to identify changes that take only 3 % of image height.
All described characteristics make our solution applicable to many areas - from psychotherapy to retail. You can see the results of our algorithm in the picture.