Enhanced Deep Learning-Based IT Employee Stress Detection for Non-Intrusive Monitoring

Employee stress is common and is one of the most critical challenges faced by employees in high-stake IT environments causing health problems and underperformance. This project is stepping into this issue by applying a model to classify people into stress or non-stress categories depends on facial images. Stress categories include anger, sadness, disgust, fear, etc. Non-stress categories include happiness, neutral activities, and surprise. To perform stress detection, we conduct Transfer Learning with Mobile Net V2 and Tensor Flow with the FER2013 pre-trained dataset. The model allows providing realtime tips to help employees overcome stress, with the option for employees to turn these suggestions on or off. Instead of eliminating the cause of stress on employees, this method seeks to place the stress-drowning activities in proper positions so that employee’s mental welfare is taken care of. Besides, the project contains a daily report with analytics and commented activities.