A trained neural network that identifies handwritten digits. For more details on how the model works and some of the methods I incorporated, please check out the Jupyter Notebook file in this repository!
I used this trained neural network in a Flask web API as the backend for my Digit Recognition Web App project.
- Neural network that identifies grey-scaled, 28 x 28 pixel images of handwritten digits
- Using the MNIST digit data set for training and validation
- Model accuracy on validation data is around 97%
- 1 input layer, 3 hidden layers, and 1 output layer
- The "mnist.csv" file was downloaded from Kaggle (training data): [link]
- MNIST images were also downloaded from Kaggle: [link]
- Matplotlib
- Pandas
- PyTorch
- Scikit-learn