ABSTRACT
Through a comparison of CNN and LSTM algorithms, the study seeks to improve the detection and classification of major sugarcane illnesses. CNN and LSTM networks were the two evaluation methods used for the two groups. 1,841 samples in total were used, of which 1,300 were used for testing and 541 for training. While the LSTM algorithm obtained 84.3830% accuracy, the CNN algorithm achieved 91.9070%. A statistically significant difference in accuracy between the two algorithms was found using an independent samples t-test, with a p-value of 0.002 (p<0.05). This suggests that because of its increased accuracy, the CNN algorithm is more useful for identifying and categorizing diseases related to sugarcane. The paper provides a solid answer for sugarcane disease diagnosis by decisively showing that CNNs outperform LSTMs in this application, offering a reliable method for classifying and detecting sugarcane diseases.
