Stress is a major health issue that significantly impacts mental stability and overall well-being. Early detection is crucial to prevent stress's adverse effects on safety and daily life. In this paper, we propose an approach to stress detection using machine learning techniques, leveraging facial landmarks extracted by Media-Pipe, a cutting-edge framework for real-time processing tasks. Analyzing these landmarks, we can accurately classify stress-related emotions based on facial expressions. We have developed a custom dataset to support this approach for classifying emotions on the human face. This ensures that the model can effectively capture and analyze the subtle details in facial expressions necessary for accurate stress detection. Traditional methods including the Face Haar Cascade algorithm have been used for similar tasks; however, Media-Pipe offers enhanced precision in facial landmark detection and superior real-time performance. Compared to these traditional methods, the approach using Media-Pipe and machine learning algorithms provides better accuracy, reduced computational cost, and shorter processing time. The facial landmarks are extracted through Media-Pipe framework subsequently, classified using algorithms such as support vector machines, k-nearest neighbors, and deep neural networks. Our results demonstrate high accuracy in stress recognition, with the deep neural network and k-nearest neighbors models achieving 99.88% and 99.94% accuracy, respectively. Remarkably, the support vector machine model achieves an impressive 100% accuracy in identifying stress-related emotions. This technology can serve as an effective tool for mental health professionals to detect depression symptoms, facilitating quicker preventive interventions.