MNIST Digit Classification API (PyTorch + FastAPI)

This project demonstrates how to deploy a custom-trained PyTorch model using FastAPI.

🔹 Model

• A SimpleCNN trained on the MNIST dataset (handwritten digits 0–9).
• Achieves ~97% test accuracy.
• Model weights are stored in mnist_cnn.pth.

🔹 Features

• Upload an image (digit) via Swagger UI or API.
• The API returns the predicted digit (0–9) and confidence score.
• Lightweight, runs on CPU or GPU.

🔹 How to Run

# Install dependencies
pip install -r requirements.txt

# Start server
uvicorn main:app --reload