DETECTION OF IMAGE-BASED CASSAVA DISEASES USING DEEP LEARNING ALGORITHM

ABSTRACT

In this paper, we proposed a novel method for detecting cassava plant diseases through deep learning models to reduce the drawbacks of traditional ways and improve food security across the globe in cassava-dependent areas. Both models were prepared on a high-resolution dataset for a variety of disease states, and other limitations linked to overfitting, although to a lesser degree, were detected. Our model exhibits a training accuracy of 98.9%, which is more than Singh et al. 87%. On the other hand, the validation accuracy is 54.35 percent, requiring the use of various sorts of generalization approaches to enhance these methodologies further.The proposed deep learning algorithms effectively detect and classify cassava diseases using image inputs, demonstrating robust performance across various disease categories, indicating potential for real-world agricultural applications. This work aims to achieve sustainable development goals 1, 2, 3, 9, and 12, and recommends modifications and enhancements to enhance generalization performance on unseen data.

Keywords: Cassava Diseases, Image-based Detection, Deep Learning, Enhanced
Convolutional Neural Networks (ECNNs)