This project explores the capabilities of deep learning in the field of handwritten digit recognition using the famous MNIST dataset. Two Jupyter notebooks have been developed: MNIST.ipynb
for the model using a Deep Neural Network (DNN) architecture and MNIST_cnn.ipynb
for the model based on a Convolutional Neural Network (CNN) architecture. This work illustrates the use of TensorFlow and Keras, highlighting the effectiveness of the DNN and CNN models for image classification.
- TensorFlow
- Keras
- NumPy
- Matplotlib
The project is organised with the following Jupyter notebooks:
- MNIST.ipynb` : Notebook for the Deep Neural Network model.
- MNIST_cnn.ipynb` : Notebook for the Convolutional Neural Network model.
To run these notebooks, make sure you have installed Jupyter as well as Python and the necessary packages. You can install the dependencies by running :
pip install tensorflow keras numpy matplotlib jupyter
To use the notebooks, run Jupyter Notebook or JupyterLab in your development environment:
jupyter notebook
Then open MNIST.ipynb
for the DNN model or MNIST_cnn.ipynb
for the CNN model and follow the instructions in each notebook to train and evaluate the models.
The notebooks contain sections dedicated to displaying model performance, including visualisations of training/test losses and accuracies, as well as confusion matrices for evaluating classification results.
This project demonstrates the effectiveness of DNN and CNN architectures applied to the recognition of handwritten digits with the MNIST dataset, using modern deep learning approaches with TensorFlow and Keras. These notebooks provide a solid basis for exploring and expanding future research in the field.