A Python project that trains a neural network to recognize hand-drawn digits between 0 and 9 using the MNIST dataset, and provides a FastAPI web service for predictions.
This project uses PyTorch to create and train a neural network for digit recognition.
The trained model can accurately classify hand-drawn digits from the MNIST dataset. Additionally, it includes a FastAPI web service that allows users to make predictions on uploaded images and view sample test images with predictions.
- PyTorch-based neural network architecture
- Training script with configurable hyperparameters
- Evaluation metrics for model performance
- Ability to save and load trained models
- FastAPI web service for making predictions on uploaded images
- Endpoint to view sample test images with predictions
- Python 3.x
- PyTorch
- FastAPI
- Pillow (PIL)
- NumPy
- Matplotlib (for visualization)
- Dataset: MNIST
- The MNIST dataset is used to train the neural network model.
- The trained model is saved and can be loaded for predictions.
- A FastAPI web service provides two main endpoints:
/predict: Accepts an uploaded image and returns the predicted digit./test_images: Generates a plot of sample test images with predictions.
- Python 3.x
- pip
- Clone the repository:
git clone https://github.com/yourusername/mnist-digit-recognizer.git cd mnist-digit-recognizer - Install dependencies:
pip install -r requirements.txt
To start the FastAPI web service, run:
python main.py
The service will be available at http://localhost:8000.
POST /predict: Upload an image file to get a prediction.GET /test_images: View a plot of sample test images with predictions.
To make a prediction on an image, send a POST request to /predict with the image file:
curl -X POST -F "file=@path/to/your/image.png" http://localhost:8000/predict
To view sample test images with predictions, open the following URL in your browser:
http://localhost:8000/test_images
This project is open source and available under the MIT License.