This repository contains code for a Convolutional Neural Network (CNN) model in PyTorch to classify Moroccan money. The code loads a dataset from Kaggle, prepares the data, builds a CNN model, trains the model, and evaluates its performance.
Make sure you have the required libraries installed:
- PyTorch
- torchvision
- matplotlib
- numpy
You can install these libraries using pip:
pip install torch torchvision matplotlib numpyThe dataset used in this project is Moroccan Money dataset. It contains images of Moroccan money that have been resized and normalized for training and testing.
To train the CNN model, run the provided code in the Moroccan_Money_classification.ipynb Jupyter Notebook. The code includes data loading, model creation, training loop, and saving the trained model.
The CNN model architecture used for this project:
- Convolutional Layer 1: 3 input channels, 6 output channels, kernel size 5x5
- Max Pooling Layer: 2x2
- Convolutional Layer 2: 6 input channels, 16 output channels, kernel size 5x5
- Fully Connected Layer 1: Input size 2704, Output size 120
- Fully Connected Layer 2: Input size 120, Output size 84
- Fully Connected Layer 3: Input size 84, Output size 11 (number of classes)
The model is trained using stochastic gradient descent (SGD) with a cross-entropy loss function. The training process is saved in a history dictionary, containing loss and accuracy values. The trained model is saved to a file named Moroccan_Money_detection.pth.
The model's performance is evaluated on a test dataset, and accuracy metrics are provided for each class.
The Moroccan_Money_classification.ipynb includes visualization of training loss and accuracy using matplotlib.
This project is for educational purposes and uses a Kaggle dataset. Credit to the Kaggle community for providing the dataset.
Enjoy experimenting with Moroccan money classification using PyTorch!
Mohamed Stifi Linkedin