Emotion-Based Smile Classifier: A Deep Learning Based Emotion Detection System

Abstract

In the modern era of digitalization, detecting human emotions from facial expressions is a crucial field of study, particularly in the context of mental health conditions. This project proposes an emotion-based smile classifier, a deep neural network built using ResNet-50 and transfer learning to classify facial expressions into four emotional categories: genuine smiles, fake smiles, neutral expressions, and anxiety/depression. The system classifies facial images with high accuracy through pre-trained convolutional layers. Through the integration of data augmentation, normalization, and fine-tuning methods, the model is made robust and generalizable. The solution is deployed with Streamlit, which delivers a friendly web interface for real-time emotion recognition and meaningful visual feedback. The goal here is to make a positive contribution to emotional consciousness and early mental health care through low-cost AI-based tools.