What kind of facial features combination will be considered high face value by more people? Recently, under the leadership of associate professor tuukka ruotsalo, scientists from the University of Helsinki and the University of Copenhagen have developed a generative confrontation neural network, generating hundreds of realistic portraits.The computer-generated images were then presented to 30 subjects, each of whom was instructed to focus more on the faces they thought were the most attractive, and to record their brain electrical activity using an electroencephalogram (EEG).
Then, the algorithm based on machine learning determines which faces generate the most activity for each person, and then determines which features these faces have in common. Based on these data, the neural network continues to generate new faces combined with these features. In a double-blind experiment, the new faces were then presented to the person along with images of many other faces. 87% chose the most attractive of the new faces - a number that should rise as technology advances.
The paper on this research was published in《IEEE Transactions in Affective Computing》In the magazine.
Source:University of Helsinki