mnist-digit-recognition-tf

🧠 MNIST Digit Classifier with CNN

A complete end-to-end Convolutional Neural Network (CNN) project using TensorFlow and Keras to classify handwritten digits from the MNIST dataset. This project is beginner-friendly and ideal for learning image classification fundamentals with deep learning.

---

📦 Dataset: MNIST

• 70,000 grayscale images of handwritten digits (0–9)
• Image size: 28×28 pixels
• Split: 60,000 for training and 10,000 for testing
• Built-in dataset in TensorFlow

---

🚀 Project Highlights

• ✅ Data loading and exploration
• 🖼️ Sample visualization and digit distribution
• 🔧 Preprocessing with normalization and reshaping
• 🏗️ CNN model building using Keras Sequential API
• 📈 Training and validation tracking
• 🔮 Model evaluation and predictions
• 📊 Confusion matrix and per-class analysis
• 🎯 Random test predictions
• 💾 Ready for further fine-tuning or saving the model

---

📊 Model Performance

• Final Test Accuracy: ~99%
• Uses dropout regularization to prevent overfitting
• Visualizes training/validation accuracy and loss

---

🧪 Requirements

• Python ≥ 3.7
• TensorFlow ≥ 2.x
• NumPy, Matplotlib, Seaborn, scikit-learn

📸 Sample Results

Install dependencies:

pip install tensorflow numpy matplotlib seaborn scikit-learn

🧠 Learnings Understanding CNN architecture

• Visualizing model performance
• Working with real image data
• Evaluating classification models