The Problem- Stress exerts significant adverse mental and physiological effects on the individual. It is imperative, therefore, to detect individuals under stress

The Solution- This project proposes a novel approach to study of non-intrusive, remote, dynamic detection and evaluation of stress in the wild with the end goal of real-time detection in online-platforms.

Our goal is to assess states of stress in individuals using internet and video platforms. To do so, we will develop computer-vision and machine-learning algorithms to track, extract, and analyze individuals' audio-visual cues and predict changes in their stress levels. The modules we will employ include classic facial motion and expressions, upper body movement, direction of gaze, pupil size and blinks, as well as skin color, through which heart rate, arousal, and micro emotions may be detected. Our algorithms will be developed on a unique video dataset of individuals in high-stress situation