Performance evaluation of different deep learning models used for the purpose of healthy and diseased leaves classification of Cherimoya (Annona Cherimola) plant

Controlling plant leaves disease helps in upholding their health. This augments the overall strength of the plant productivity both in terms of higher quality and quantity. Recently, with the expansion of high-end computing devices, Artificial Intelligence (AI) techniques have allied in almost all applications. In agriculture, AI has been used in crop surveillance, soil health monitoring, nutrient deficiency estimation, flower and fruits quality assessment, weed estimation, and crop health diagnosis among all. Therefore, in this work, five deep learning approaches, namely EfficientNet, MobileNetV2, BiT, EANet, and Swin Transformers, have been used for the classification of healthy and diseased leaves of Cherimoya plant. EANet model with precision = 0.9929%, recall = 0.9536%, F1-score = 0.9846%, training accuracy = 0.9775%, and testing accuracy = 0.9689% achieved superior performance among all. We believe that the proposed work can be very useful in providing an accurate and timely plant disease diagnosing system.