Classification of Plant Diseases using DenseNet 121 Transfer Learning Model

A significant risk to agricultural output and food security is posed by plant diseases. Plant diseases may be identified and diagnosed using traditional methods, but these procedures can be laborious and complicated. As the process is complicated, it is easy for a plant disease to be misdiagnosed which can lead to further damage to the plant. Thus, to overcome the problems faced by traditional methods, diagnosing a plant disease in early stages, and increase the reliability of the diagnosis obtained, deep learning can be used for classification of plant diseases. This paper aims to increase agricultural productivity and ensure food security by preventing crops from being affected by plant diseases through timely diagnosis and accurate classification. This would further develop the goal of sustainable food production in the world. In this paper, the deep transfer learning model DenseNet121 is used to detect and classify plant diseases for pepper bell, potato, and tomato plants with multiple fine-tuning layers for improved and refined results. An accuracy of 97.38% was achieved with the model proving that deep learning methods can be used for the effective diagnosis and classification of plant diseases.