Emotion Recognition Using Convolutional Neural Networks

Emotion Recognition Using Convolutional Neural Networks (CNN) Developed and deployed a deep learning system for facial emotion recognition classifying images into 7 emotions using a custom CNN architecture. Built a 3-block CNN architecture with Batch Normalization, Progressive Dropout, and Global Average Pooling Applied data preprocessing, normalization, and real-time data augmentation (flip, rotation, zoom) Implemented Adam optimizer, learning rate scheduling (ReduceLROnPlateau), and L2 regularization to reduce overfitting Achieved 67.05% test accuracy, approaching human-level performance on FER-2013 benchmark Performed detailed evaluation using confusion matrix, classification report, learning curves analysis Deployed the model as a full-stack web application using Flask (backend) and HTML/CSS/JS (frontend) with live inference Technologies: Python, TensorFlow/Keras, OpenCV, Flask, NumPy, Matplotlib, Google Colab