Emotion Classifier – CNN & RNN

This repository contains two deep learning–based emotion classification systems:

1. CNN-based Facial Emotion Recognition (Image-based)
2. RNN-based Emotion Classification (Text-based)

The CNN model is production-ready, Dockerized, and deployable on Docker & Kubernetes.

---

📁 Project Structure

Emotion_Clasifier-RNN-CNN-
│
├── CNN/
│   ├── app.py                  # Flask API for CNN inference
│   ├── model.h5                # Trained CNN model (emotion classification)
│   ├── Dockerfile              # Docker configuration for CNN service
│   ├── requirements.txt        # Python dependencies
│   ├── cnn-deployment.yaml     # Kubernetes Deployment
│   ├── cnn-service.yaml        # Kubernetes Service (ClusterIP)
│   ├── test.py                 # Client-side test script
│   └── Image Classifier.ipynb  # CNN training & experimentation notebook
│
├── RNN/
│   ├── model.h5                # Trained RNN model (text emotion classification)
│   ├── Predection_UI.py        # Local UI / script for prediction
│   ├── TextClassifier1.ipynb   # RNN training notebook
│   ├── training.csv            # Training dataset
│   └── README.md               # RNN-specific documentation

---

🧠 CNN – Facial Emotion Recognition

A Convolutional Neural Network (CNN) trained to classify facial emotions from images. The model is exposed as a Flask REST API and packaged as a Docker container.

🎭 Supported Emotions

• Angry
• Disgusted
• Fearful
• Happy
• Neutral
• Sad
• Surprised

---

🐳 Docker Image (CNN)

The CNN service is available as a public Docker image:

🔗 Docker Hub Repository https://hub.docker.com/repository/docker/brijesh944/emotion_cnn/general

▶️ Run with Docker

docker run -p 5000:5000 brijesh944/emotion_cnn

Access:

http://localhost:5000

---

☸️ Run CNN with Kubernetes (Docker Desktop / Minikube)

Deploy

kubectl apply -f cnn-deployment.yaml
kubectl apply -f cnn-service.yaml

Verify

kubectl get pods
kubectl get svc

Inside the cluster, the service is accessible via:

http://cnn:5000

From local machine:

kubectl port-forward svc/cnn 5000:5000

---

📡 CNN API Endpoint

POST /predict

Request (JSON):

[[[[...image pixel values...]]]]

Response:

{
  "predicted_class": 3
}

---

🧪 Testing the CNN API

Use the provided script:

python test.py

Or send a request using requests:

requests.post("http://localhost:5000/predict", json=data)

---

🧠 RNN – Text Emotion Classification

The RNN module focuses on text-based emotion detection using sequential models.

Features

• Trained RNN model (model.h5)
• Notebook for training & evaluation
• Local prediction UI / script
• CSV-based training dataset

Note: The RNN module is currently designed for local execution, not containerized.

---

🛠️ Tech Stack

• Python 3
• TensorFlow / Keras
• Flask
• NumPy
• Docker
• Kubernetes
• Jupyter Notebook

---

🎯 Use Cases

• Facial emotion recognition demos
• Text-based emotion analysis
• ML inference microservices
• Docker & Kubernetes learning
• MLOps and deployment practice

---

⚠️ Notes

• The CNN service binds to 0.0.0.0:5000 for container compatibility
• Models are intended for educational and prototype purposes
• Not optimized for large-scale production inference

---

👤 Author

Brijesh Computer Science Engineer | AI & Deep Learning Enthusiast

---

If you plan to extend this project:

• Add RNN Docker support
• Introduce model versioning
• Add Ingress / LoadBalancer
• Implement CI/CD for ML pipelines

🚀