🧠 Handwritten Digit Recognition (MNIST)

Deep learning project that classifies handwritten digits (0–9) using neural networks and CNNs on the MNIST dataset.

---

📁 Project Structure

Handwritten-Digit-Recognition/
├── data/
│   ├── mnist_train.csv
│   └── mnist_test.csv
├── notebooks/
│   └── handwritten_digit_recognition.ipynb
├── models/
│   └── mnist_cnn_model.h5
├── images/
│   ├── training_plot.png
|   ├── training_plot2.png
│   ├── confusion_matrix.png
│   └── predictions.png
└── README.md

---

⚙️ Setup

pip install pandas numpy matplotlib seaborn tensorflow scikit-learn

---

🚀 Run Notebook

git clone https://github.com/Laithwm/Handwritten-Digit-Recognition.git
cd Handwritten-Digit-Recognition
jupyter notebook notebooks/handwritten_digit_recognition.ipynb

---

📊 Results

• Dense Neural Network → ~99.2% accuracy
• CNN Model → 99.33% test accuracy
• Most errors occurred between visually similar digits (e.g., 4 vs 9).
• Precision, recall, and F1-scores were all above 0.99.

---

🧠 Skills

• Data preprocessing & normalization
• Neural network & CNN modeling
• Model evaluation & visualization

---

Author: Laith Waqas

Convolutional architectures are crucial for image-based classification tasks, as evidenced by the CNN model's overall superior ability to capture local image features.

📍 Dublin, Ireland
💼 LinkedIn