Real-Time Detection and Analysis of Facial Movement Disorders Using MediaPipe and OpenCV: A Streamlit-Based Framework

This study introduces an innovative framework for Facial Movements Disorders (FMD), including eyelid movement, and mouth droopy. The system operates in real-time through front-facing laptop cameras, uploaded videos, and images. The approach involves initializing the MediaPipe face landmark model, preprocessing frames/images using OpenCV, and extracting landmarks for identifying facial points. By analyzing distances, predefined thresholds, and movements are classified. We implemented a user-friendly web interface using the Streamlit library to empower users to use the implemented model easily. The system was rigorously evaluated on a labeled dataset. Evaluation metrics (accuracy, precision, recall, F1 score) were calculated. Results indicated a high accuracy of 93% for eyelid status, and 95% for mouth droopy corner.