Handwritten Digit Recognition with Convolutional Neural Networks

---

📌 Overview

This repository provides an end-to-end Convolutional Neural Network (CNN) pipeline for classification of handwritten digits (0–9) on the standard MNIST benchmark.

The project investigates the impact of network depth, convolution filter sizes, spatial subsampling (max-pooling), and dropout regularization on training stability, test accuracy, and generalization capability.

---

Key Components:

• Feature Extraction: Stacked convolutional layers capturing hierarchical spatial features (low-level edges and corners progressing to stroke geometries).
• Dimensionality Reduction: Non-overlapping max-pooling layers enforcing spatial translation invariance.
• Overfitting Prevention: Dropout regularization to mitigate co-adaptation of hidden unit activations.
• Diagnostics: Loss convergence curves, per-class precision/recall, and confusion matrix error analysis.

---

🚀 Getting Started

git clone https://github.com/dheerajkumar2005/Handwritten-Digit-Recognition-CNN.git
cd Handwritten-Digit-Recognition-CNN

pip install numpy matplotlib seaborn jupyter scikit-learn tensorflow

jupyter notebook cnn-handwritten-digit-recognition.ipynb