EmotionAI: Facial Emotion Detection System

Overview

Facial Emotion Detection System is an open-source project designed to identify human emotions from images and real-time video streams using advanced Convolutional Neural Networks (CNNs). The project is modular, containerized using Docker, and made accessible for global use.

---

Features

1. Backend: A RESTful API built with FastAPI for image-based emotion detection.
2. Frontend: A static web interface for uploading images and viewing predictions.
3. Real-Time Video Detection: Real-time emotion detection from video streams using OpenCV.
4. Dockerized Deployment: Pre-built Docker containers for easy local and remote usage.

---

Folder Structure

Emotion-Detection-Using-CNN-and-FER-2013-Dataset/
├── backend/                      # Backend service
│   ├── main.py                   # FastAPI implementation
│   ├── requirements.txt          # Dependencies
│   ├── weights/                  # Pretrained model weights (not in repo)
│   │   └── ResNet50_final_weights.weights.h5
│   └── Dockerfile                # Dockerfile for backend
├── frontend/                     # Frontend service
│   ├── index.html                # Main HTML file
│   ├── app.css                   # Styling
│   ├── app.js                    # Frontend logic
│   └── Dockerfile                # Dockerfile for frontend
├── video_prediction/             # Real-time video prediction
│   ├── video_emotion_detector.py # Video detection script
│   ├── haarcascade_frontalface_default.xml # Haar cascade file
│   ├── requirements.txt          # Dependencies
│   └── Dockerfile                # Dockerfile for video service
├── docker-compose.yml            # Docker Compose configuration
└── README.md                     # Comprehensive documentation

---

How to Use

Open-Source Docker Containers

This project is containerized and available as pre-built Docker images for global use. Follow these steps to pull and run the containers:

1. Pull the Docker Images

Run the following commands:

docker pull your-dockerhub-username/emotion-detection-backend
docker pull your-dockerhub-username/emotion-detection-frontend
docker pull your-dockerhub-username/emotion-detection-video

2. Run the Containers

1. Backend:

   docker run -d -p 8000:8000 your-dockerhub-username/emotion-detection-backend

   • Accessible at: http://localhost:8000/docs

2. Frontend:

   docker run -d -p 3000:3000 your-dockerhub-username/emotion-detection-frontend

   • Accessible at: http://localhost:3000

3. Video Prediction:

   docker run -it --rm your-dockerhub-username/emotion-detection-video

---

Local Setup (Without Docker)

1. Clone the Repository

git clone [email protected]:Muhammad-Sheraz-ds/Emotion-Detection-Using-CNN-and-FER-2013-Dataset.git
cd Emotion-Detection-Using-CNN-and-FER-2013-Dataset

2. Install Dependencies

Backend:

cd backend
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cd ..

Frontend: No installation required. Open frontend/index.html in your browser.

Video Prediction:

cd video_prediction
pip install -r requirements.txt
cd ..

3. Run the Application

1. Start Backend:

   cd backend
   uvicorn main:app --reload
   cd ..

2. Run Video Prediction:

   cd video_prediction
   python video_emotion_detector.py

3. Open the frontend index.html in your browser.

---

Accessing the Model

Backend API

The backend is deployed as a RESTful API and can be accessed via Swagger UI:

• http://localhost:8000/docs

Endpoint: /predict-image/

• Method: POST
• Description: Predict emotion from an uploaded image.
• Input: An image file (multipart/form-data).
• Output: JSON response with the detected emotion.

Example cURL Command:

curl -X POST "http://localhost:8000/predict-image/" \
-H "accept: application/json" \
-H "Content-Type: multipart/form-data" \
-F "file=@path/to/image.jpg"

Response:

{
  "emotion": "Happy"
}

Frontend

• Open the frontend in your browser (http://localhost:3000).
• Upload an image and view the predicted emotion.

Real-Time Video Prediction

Run the video prediction container:

docker-compose up video_prediction

---

Model Weights

The pretrained model weights are not included in the repository due to size limitations. To use this project:

1. Download the weights from the provided Google Drive link.
2. Place the downloaded ResNet50_final_weights.weights.h5 file in the backend/weights/ directory.

Emotion-Detection-Using-CNN-and-FER-2013-Dataset/
├── backend/
│   ├── weights/
│   │   └── ResNet50_final_weights.weights.h5

---

Testing the Application

1. Use Swagger UI to test backend endpoints (http://localhost:8000/docs).
2. Verify the frontend by uploading images and viewing predictions.
3. Test real-time video emotion detection by running the video module.

---

Future Enhancements

• Cloud deployment for global accessibility.
• GPU acceleration for real-time video prediction.
• Improved frontend with mobile responsiveness.

---

License

This project is licensed under the MIT License.

---

Let me know if this aligns with your requirements! 😊