Stress is a part of each and everyone’s life, and we all have to face it at one time or another when things get tougher. It doesn’t only affect our health and mood, but also the way we interact with others. Early detection will help people reduce stress since one can overcome the problem before things get out of control. The automated model presented in this paper can detect stress from human facial expression. The key emotions that this model used to assess stress levels are anger, fear, disgust, surprise, neutrality, sadness, and happiness. Using deep learning techniques, mainly Convolutional Neural Networks (CNNs), the proposed model accurately predicts these emotions. The CNN model acquired a prediction accuracy of 97.2%, while other machine learning models were also used to assess the performance. Based on the intensity of each observed expression, this model determines the stress level.