ARCHIVES

Our Work address a deep learning-based approach on developing a deep learning-based system for automatic facial emotion recognition from video data. The proposed approach processes video input sequentially by extracting frames, detect-ing faces using the Multi-task Cascaded Convolutional Neural Network (MTCNN), and classifying each face into predefined emotion categories through a convolutional neural network model. A user-friendly web application built with Flask enables video upload, real-time inference, and extraction of segments corresponding to specific emotions. The model is trained and tested on the FER2013 dataset to ensure reliable performance. Experimental results indicate that the system effectively rec-ognizes facial expressions in video streams, achieving results comparable to established benchmarks. Additionally, a confusion matrix is utilized to evaluate classification performance across different emotion classes.