The Role of Deep Learning in Transforming Cassava Leaf Disease Diagnosis

Abstract

The cultivation and output of cassava, the most important crop in the Philippines, are greatly endangered by a number of illnesses. Because cassava is essential for food security, it is crucial to identify these illnesses in their early stages before they worsen. The approaches utilised in disease identification are based on the expert knowledge of fuzzy domain products and the framers’ experience. This paper proposes an automated system for illness identification using pretrained neural network models and deep learning, with inspiration from convolutional neural network techniques. The system would deal with fuzzy input. Consequently, the suggested neural network systems automatically detected red leach and anthocyanin. In this case, a model was trained using transfer learning on the EfficientnetB3 configuration, which had 10,791,220 learnable parameters, in order to deal with fuzzy data for cassava disease diagnosis. With the suggested data model, which incorporates the same, the accuracy reached was 89.90%, surpassing the 50.46% baseline proportionate chance consideration. In spite of data imbalances, the neural network classifier still manages to identify cassava leaf disease with remarkable accuracy. Improved crop yields aren't the only thing that farmers and agricultural researchers can do with our findings. In order to avoid crop failure, our method of disease detection may automate data collecting for disease diagnosis. Our findings lead us to suggest the following studies as potential avenues for further study into plant disease detection at a larger scale.

Topics & Concepts