Novel semi-supervised AI method

Due to notable differences between human beings, analysis of human behaviour does not work in "one fits all" way. Instead, AI methods should be adapted to each end user, using small number of labels. VTT has developed a semi-supervised method that requires fewer labels than was to date required in other stress detection studies. This method allowed to detect not only stress, but also certain stressors and hence can be potentially applicable to other areas where AI adaptation to differences in humans and/ or contexts is required and data sizes are not sufficiently big for deep learning.

AI methods for semi-supervised classification and active learning

Currently, AI methods run on Windows.

Companies in need to understand human behaviour

IPR, commercial projects