Early Detection of Crop Diseases using Deep Learning Techniques

Empowering Cocoa Farmers

Abstract

Cocoa, a vital component of chocolate production, faces significant threats from diseases that pose severe risks to crop yield and the livelihoods of cocoa farmers. Traditional disease detection methods, reliant on subjective visual inspections by farmers, lack precision and hinder timely intervention. To address these challenges, our goal is to utilize deep learning techniques to revolutionize disease identification in cocoa crops in Africa. We plan to achieve this by developing and training the YOLOv5 model on a diverse and recently published African dataset of cocoa crop images showcasing various diseases, providing farmers with a practical tool capable of accurately identifying areas suspected to be affected by diseases. The proposed model will be deployed through Flask and mobile applications, enabling real-time disease detection and immediate feedback for farmers. The expected output of our project is a robust classification model integrated into user-friendly applications accessible via mobile devices. Farmers can capture images of their crops, receive instant analysis, and get actionable feedback on disease presence or absence. This approach enables prompt intervention and mitigation strategies, ensuring the sustainability of cocoa production. Our next steps involve extensive testing and validation of the model using a diverse range of cocoa crop images. We will also collaborate with cocoa farming communities to obtain user feedback, refine the model, and address specific regional considerations. Ultimately, we aim to scale the application to cater to larger cocoa-dependent regions, contributing to the advancement of sustainable agriculture practices.