Smart Agriculture: Predicting Diseases in olive using Deep Learning Algorithms

¹ Department of CS & IT, Kalinga University, Raipur, India
² Research Scholar, Department of CS & IT, Kalinga University, Raipur, India

^* Corresponding author: ku.frahman@kalingauniversity.ac.in

Abstract

The olive industry is both economically and culinary important, and there are numerous diseases threatening it. Typically, manual inspections and lab analyses for detecting and managing disease in olive cultivation are time consuming and subject to delay. In this study, olive tree diseases are managed by improving prediction and management using a deep learning algorithm, with the aim of efficiency and accuracy of the process. Utilizing a large number of olive leaf image data and corresponding environment factors, we developed and evaluated Convolutional Neural Networks (CNNs) and Long Short Time Memory networks (LSTMs). We trained these models to spot early signs of disease from the visual symptoms and predict potential outbreaks from the given data. Using image data together with temperature, humidity, soil conditions, etc. our method offers a holistic view on disease dynamics in olive cultivation. The results show that traditional methods are no longer competitive against other types of models, in both accuracy and processing time | detection accuracy | processing speed. Our best model was able to identify diseased leaves with an accuracy of 94% and could be used for real time disease monitoring and early therapy. The level of precision here is so high, that farmers can take timely action, that can stop the spread of diseases and save the loss in the economy. These implications are profound and point towards a scalable and cost effective solution on disease prediction on olive plantations. The application of these advanced practices by the agricultural stakeholders will enable better disease management practices that aid in more sustainable and productive olive farming. This work will be taken further towards modeling more refined versions of the models applied to additional crops and using more diverse sources of data to increase the resilience of agriculture to disease threats.