mnist-digit-classifier

🧠 MNIST Digit Classifier (Deep Learning Project)

A simple but powerful deep learning model that classifies handwritten digits (0–9) using the famous MNIST dataset.

This project is beginner-friendly and built using TensorFlow and Keras, achieving over 97% accuracy on test data. It's perfect for understanding how Artificial Neural Networks work.

---

📌 Project Highlights

• 📊 97%+ accuracy on test images
• 🔢 Trained on 60,000 handwritten digit images
• 🧠 Uses a basic Artificial Neural Network (ANN)
• 🖼 Visualizes predictions vs real labels
• ⚡ Built entirely in Google Colab
• 🌎 Easy to understand and modify

---

🚀 Tech Stack

Tool	Purpose
Python	Programming language
TensorFlow/Keras	Deep learning framework
Matplotlib	Visualizing predictions
Google Colab	Free GPU-powered training

---

🧠 How It Works

1. Load the MNIST dataset (built into Keras)
2. Normalize the data to improve training speed
3. Build an ANN with:
   • Input Layer: 28×28 image → Flattened
   • Hidden Layer: 128 neurons with ReLU
   • Output Layer: 10 neurons (softmax)
4. Train the model over 5 epochs
5. Evaluate performance on test data
6. Visualize predictions

---

📸 Preview

---

🧪 Example Prediction

Prediction: 7
Actual Label: 7
Confidence: 99.2%

---

🗂 Folder Structure