Soybean Leaf Disease Classification using MobileNetV2 and EfficientNetB0

Soyabean is an important crop used for food, animal feed, and industrial application. Disease symptoms on leaves can reduce crop health and productivity, making timely identification valuable. This paper presents a deep-learning workflow for classifying soybean leaf images into six dataset categories: Bacterial Blight, Ferrugun, Healthy_pic, Powdery Mildew, Sudden Death Syndrome and Yellow Mosaic. Two transfer-learning architectures; MobieNetV2 and EfficientNetB0, were tested on a prepared data set of 957 images. The data were organized into training, validation, and tests. They consisted of 654, 157, and 146 images in the subsets, respectively. MobileNetV2 test accuracy (recorded) was 99.32%. The new EfficientNetB0 evaluation claimed a 100% accuracy; however, this result is dependent on the following reproduction due to different previous assessments. The paper addresses the issue of preprocessing, model design, model analysis, and model implementation, evaluation of metrics, limitations, and a proposed bilingual prediction interface. The findings are supporting evidence and do not set equivalent performance on unstimulated field conditions only does so for the selected test split..