📝 Project Summary: This project aims to automatically detect emotions from facial images using CNNs. The model classifies expressions like 😃 happy, 😢 sad, 😠 angry, etc., benefiting affective computing and human-computer interaction.
📚 Literature Review:
📄 Paper 1: "Deep Ensemble of Fine-tuned CNNs for Facial Expression Recognition in the Wild" (2022) 🎯 Accuracy: 88.5% 🤩 Pros: Deep ensemble techniques 🤔 Cons: Computationally intensive.
📄 Paper 2: "Facial Expression Recognition via Deep Graph Convolutional Networks" (2023) 🎯 Accuracy: 85.2% 🌟 Pros: Graph CNNs capture spatial dependencies.
🧠 Model Architecture: Custom CNN with convolutional, max-pooling, and fully connected layers. 🚀 Hyperparameter tuning to optimize performance.
🗂️ Dataset: Expression in-the-Wild (ExpW) dataset 📊 91,793 facial images with emotions labeled.
📂 Data Division: Training: 70% ⚙️ Validation: 15% ⚖️ Test: 15%
⚙️ Hyperparameter Tuning: Grid search and cross-validation 🎛️
📈 Results and Evaluation: Metrics: Accuracy, precision, recall, F1-score 📊 Confusion matrix for deeper insights.
📊 Analysis of Results: Good: High accuracy, balanced precision-recall ⭐ Bad: Misclassifications, low precision/recall for some emotions.
🔮 Improvement Strategies: Data Augmentation 🔄 Transfer Learning 🔄 Ensemble Methods 🔗