🔢 Handwritten Digit Recognition — MLOps Project

📌 Project Overview

This project is a Handwritten Digit Recognition System built using Machine Learning and deployed as an interactive Streamlit web application.

The model is trained using the Scikit-learn Digits Dataset and uses a Support Vector Classifier (SVC) to classify handwritten digits from 0 to 9.

The project also demonstrates fundamental MLOps practices, including:

• Machine Learning model training
• Image preprocessing
• Model serialization
• Streamlit application
• Logging
• Docker containerization
• Git/GitHub version control
• Model evaluation
• Hyperparameter configuration

---

🚀 Features

• ✍️ Upload a handwritten digit image
• 🖼️ Image preprocessing and normalization
• 🔢 Predict digits from 0 to 9
• 🤖 SVC-based classification
• 📊 Classification report
• ⚙️ Hyperparameter configuration
• 📝 Application logging
• 🐳 Dockerized application
• 🌐 Interactive Streamlit UI
• 📦 Saved trained model using Pickle

---

🧠 Machine Learning Model

Algorithm

Support Vector Classifier (SVC)

SVC is a supervised machine learning algorithm that finds an optimal decision boundary between different classes.

For this project, the classifier is trained to recognize:

0  1  2  3  4  5  6  7  8  9

Dataset

The project uses the built-in:

Scikit-learn Digits Dataset

Dataset characteristics:

• Number of classes: 10
• Image size: 8 × 8
• Total features: 64
• Pixel values: approximately 0–16
• Classes: digits 0–9

---

🔄 Machine Learning Pipeline

              Digits Dataset
                    │
                    ▼
             Data Preparation
                    │
                    ▼
             Train / Test Split
                    │
                    ▼
             Feature Processing
                    │
                    ▼
             SVC Model Training
                    │
                    ▼
            Hyperparameter Tuning
                    │
                    ▼
              Model Evaluation
                    │
                    ▼
               model.pkl
                    │
                    ▼
             Streamlit Application
                    │
                    ▼
             Docker Container

---

🖼️ Image Preprocessing

For uploaded real-world digit images, the application performs preprocessing before sending the image to the model.

The preprocessing pipeline includes:

Input Image
    ↓
Grayscale Conversion
    ↓
Image Inversion
    ↓
Resize to 8 × 8
    ↓
Pixel Normalization
    ↓
Scale to 0–16
    ↓
Flatten to 64 Features
    ↓
SVC Prediction

This preprocessing is important because the original Scikit-learn Digits Dataset contains small 8 × 8 images with pixel values in a similar range.

---

📊 Model Evaluation

The trained SVC model is evaluated using classification metrics such as:

• Accuracy
• Precision
• Recall
• F1-score
• Support

Classification Report

The generated classification report is available in the public directory.


---

⚙️ Hyperparameters

The SVC model was trained and evaluated using selected hyperparameters.

The hyperparameter results/configuration are available below:

The hyperparameters can be modified and evaluated to find a better-performing model.

---

🛠️ Tech Stack

Technology	Purpose
Python	Programming language
NumPy	Numerical operations
Pandas	Data manipulation
Scikit-learn	Machine Learning
SVC	Classification algorithm
Pillow	Image processing
Matplotlib	Visualization
Streamlit	Web application
Pickle	Model serialization
Docker	Containerization
Git	Version control
GitHub	Source code management

---

📁 Project Structure

Digits/
│
├── .dockerignore
├── .gitignore
├── Dockerfile
├── README.md
├── requirements.txt
│
├── main.py
├── config.py
├── model.pkl
│
├── core/
│   ├── __init__.py
│   ├── digits.py
│   └── logger.py
│
├── public/
│   ├── classification_report.png
│   └── hyperparameters.png
│
└── logs/
    └── app.log

---

💻 Local Installation

1. Clone the Repository

git clone <YOUR_GITHUB_REPOSITORY_URL>

Move into the project directory:

cd Digits

---

2. Create Virtual Environment

python3 -m venv venv

Activate the environment on Linux/macOS:

source venv/bin/activate

For Windows:

venv\Scripts\activate

---

3. Install Dependencies

pip install -r requirements.txt

---

4. Run the Streamlit Application

streamlit run main.py

The application will be available at:

http://localhost:8501

---

🐳 Run with Docker

This project is containerized using Docker.

Build Docker Image

docker build -t digits-app .

Run Docker Container

docker run -p 8501:8501 digits-app

Then open:

http://localhost:8501

Docker Architecture

Dockerfile
     │
     ▼
Docker Image
     │
     ▼
Docker Container
     │
     ▼
Streamlit Application
     │
     ▼
Port 8501

---

📝 Logging

The application includes logging functionality for tracking application events.

Logs can contain information such as:

• Model loading
• Image preprocessing
• Prediction status
• Prediction result
• Errors and exceptions

Example:

2026-09-15 16:00:00 - INFO - Digits model loaded
2026-09-15 16:00:05 - INFO - Starting image preprocessing
2026-09-15 16:00:05 - INFO - Image preprocessing completed
2026-09-15 16:00:06 - INFO - Prediction result: 7

Application logs are stored in:

logs/app.log

---

🔮 Future MLOps Improvements

The project can be extended into a complete production-oriented MLOps pipeline.

Planned improvements include:

• GitHub Actions CI/CD
• Automated unit testing
• MLflow experiment tracking
• DVC for dataset/model versioning
• Model Registry
• Automated Docker image publishing
• Cloud deployment
• Model monitoring
• Data drift detection
• Model performance monitoring
• Automated model retraining

---

📈 MLOps Roadmap

Machine Learning
       │
       ▼
Model Serialization
       │
       ▼
Streamlit Application
       │
       ▼
Logging
       │
       ▼
Docker
       │
       ▼
Git / GitHub
       │
       ▼
CI/CD ────────────────┐
       │              │
       ▼              ▼
   MLflow             DVC
       │              │
       └──────┬───────┘
              ▼
        Model Registry
              │
              ▼
          Deployment
              │
              ▼
         Monitoring
              │
              ▼
         Retraining

---

👨‍💻 Author

Muhammad Haseeb Raza

Machine Learning Engineer

📧 Email: hasiraza511@gmail.com

Areas of Interest

• Machine Learning
• Deep Learning
• Computer Vision
• MLOps
• Model Deployment
• Python
• Data Science

---

⭐ Project

If you find this project useful, feel free to ⭐ star the repository and explore the code.

---

📄 License

This project is created for educational and portfolio purposes.