PLANT LEAF DISEASE CLASSIFICATION USING DEEP LEARNING

Abstract

The prevalence of plant diseases is in fact one of the main factors that lower the quality and quantity of agricultural products. The diseases keep emerging in the leaves of the plants with the development in plant structure and change in cultivation methods. Usually, the diseases first attack the leaves and then spread to the whole plant; hence the variety and yield of the crops that can be grown get highly influenced. Plant diseases are, in fact, one of the leading prevailing points of attacks on the global food supply and funds. This work has developed a system using EfficientNetV2 for plant leaf disease classification. The model has been trained on the PlantVillage dataset, which now contains 61,486 manually labeled images showing 14 different classes of healthy or unhealthy crop leaves and categorized over 39 distinct classes. Extensive testing and comparison showed that the model properly identifies plant leaf diseases. This all is going to draw the conclusions able to revolutionize the strategy for disease detection and control in plants. The experimental results revealed that the EfficientNetV2 model was able to give an accuracy of 99.40% in training and 99.24% in testing, suggesting its high effectiveness for early diagnosis of leaf diseases. In addition to that, with the implementation of deep learning and lately designed EfficientNetV2, it offers an effective way to timely disease detection for the improvement of agricultural practice, which aims at global food security.