Identifying and classifying plant disease using resilient LF-CNN

Abstract

Food security is an important factor in maintaining the livelihood of people around the world. Plant biosecurity mainly deals with analyzing and managing the health of the plant. The biosecurity measures help in reducing the transmission of disease in plants. Environmental factors have a direct influence on determining the growth, stability, and resistance over a variety of diseases. Plants are highly vulnerable to seasonal diseases and the progression increases over time under different environmental conditions. So, it is indeed important to address the problem of protecting the plants from heterogeneous diseases. Many computational techniques have been proposed to early detect the plant disease to protect the crops from devastation. But, the performance of the existing system needs improvement to enhance the predictive ability of the model in challenging situations. In this paper, an effective loss-fused convolutional neural network model is proposed to identify the plants affected with disease of its own type. This system combines the advantages of two different loss functions thereby makes better prediction. The diseases were classified based on the features extracted from the plant leaves in the final layer of the model. The dataset used to perform this experiment is accessed from Plant Village Database. This system attained 98.93% accuracy on discriminating the affected samples over the unaffected one. The result obtained through this model proves its efficacy on the classification of disease affected leaf samples over other existing methodologies.