Egyptian Currency Classification using Custom CNN (From Scratch)

I developed a complete Computer Vision project that classifies Egyptian currency notes from the new 2023 series using a Convolutional Neural Network (CNN) built entirely from scratch, without using any pre-trained or transfer learning models.

The model was implemented using Python and TensorFlow/Keras and trained on a real dataset of Egyptian banknotes. The system is capable of accurately recognizing multiple currency denominations and achieved high classification performance.

Key Features of the Project

•Designed and implemented a custom CNN architecture from scratch

•Applied data augmentation to improve generalization and prevent overfitting

•Used Batch Normalization, Dropout, and Learning Rate Scheduling

•Implemented Early Stopping to optimize training

•Achieved high classification accuracy (≈93%+)

•Generated training/validation accuracy and loss curves

•Evaluated the model using:

•Confusion Matrix

•Classification Report (Precision, Recall, F1-score)

•Tested the model on unseen images

•Saved trained model and weights for deployment