001kazeem/yoruba-character-recognition-cnn

A CNN-based recognition system for Yoruba characters with 5,000 labelled images. The model achieved 89.17% training accuracy and 88.08% validation accuracy, demonstrating the potential of deep learning for low-resource language technology. # \# Yoruba Character Recognition Using Deep Learning # # \## Overview # # This project presents a Deep Learning-based approach for recognizing Yoruba characters using Convolutional Neural Networks (CNNs). # The research explores the application of Artificial Intelligence techniques to low-resource language technology by developing a system capable of classifying Yoruba character images accurately. # Yoruba is one of the widely spoken African languages, yet digital language resources remain limited compared to high-resource languages. This project demonstrates how Machine Learning can contribute to language preservation, accessibility, and intelligent language systems. # # \## Research Objective # # The main objective of this project is to develop an automated system capable of recognizing Yoruba characters from image data using deep learning techniques. # # The project aims to: # \- Develop a CNN-based character recognition model. # \- Build a complete machine learning pipeline from data preparation to evaluation. # \- Investigate AI applications for low-resource language technologies. # \- Provide a foundation for future research in NLP and language models. # # \## Dataset # # The dataset used in this research contains approximately 5,000 labelled Yoruba character images. # # Dataset details: # \- Dataset size: 5,000 images # \- Data type: Image-based character samples # \- Task: Multi-class classification # # Preprocessing steps included: # \- Image resizing # \- Pixel normalization # \- Data augmentation # \- Training and validation splitting # # \## Methodology # The project followed a machine learning pipeline consisting of: # # 1\. Data collection and annotation # 2\. Data preprocessing # 3\. CNN model development # 4\. Model training # 5\. Performance evaluation # # The model was trained using Convolutional Neural Network techniques because CNNs are effective for extracting visual features from image-based data. # …