Facial Emotion Detection

This deep learning system recognizes seven human emotions — happy, sad, angry, surprise, fear, disgust and neutral — from live webcam video in real time. OpenCV first detects and crops each face using Haar cascades; a convolutional neural network trained on the FER2013 dataset of about 35,000 labelled facial images then classifies the expression, achieving around 65% benchmark accuracy with smooth 15-25 FPS inference on a standard laptop CPU. On-screen labels and emotion-confidence bars update frame by frame, and results can be logged for session analytics. Real-world applications include engagement-aware e-learning, driver stress and drowsiness monitoring, interview analytics and mental-health screening support. Built with Python, TensorFlow/Keras and OpenCV, this is a flagship BTech CSE artificial intelligence project that combines computer vision, deep learning and real-time deployment.