Emotion Detection Using CNN and FER-2013

🚀 Emotion Detection (April 2025) Deep Learning project built with Python, TensorFlow, Keras, VGG16, ResNet50v2, OpenCV, and Gradio.

This project trains a Convolutional Neural Network (CNN) on the FER-2013 dataset to classify 7 human emotions and deploys a real-time emotion recognition system using OpenCV and Gradio.

✨ Features

📊 CNN Model trained on FER-2013 with data augmentation, dropout, and class weighting.

🎯 Achieved 66% validation accuracy.

🖼️ Supports 7 emotions: Angry, Disgust, Fear, Happy, Sad, Surprise, Neutral.

🎥 Real-time detection from webcam using OpenCV.

🌐 Interactive interface powered by Gradio for user-friendly predictions.# Emotion_detection Results

Validation Accuracy: 66%

Techniques used:

Data Augmentation

Dropout Regularization

Class Weighting for imbalance

📷 Demo

Gradio Interface:

Real-time Webcam Prediction:

🔮 Future Improvements

Improve accuracy with EfficientNet / Vision Transformers (ViT).

Deploy as a web app or mobile app.

Train with larger datasets for robustness.

📚 Tech Stack

Python 3.9+

TensorFlow / Keras

OpenCV

Gradio

VGG16, ResNet50v2 pretrained models