Efficient Plant Leaf Disease Detection Using a Customized Convolutional Neural Network

Abstract

Plant leaf disease diagnosis is a complex and very crucial task for agriculture due to its significant role in crop production, quality, and safety. Numerous machine learning-based methods have been proposed for the same. However, these methods often suffer from overfitting and the efficiency of the performance. A customized convolutional neural network is proposed to detect abnormal plant leaves efficiently. The proposed model uses data augmentation and handles overfitting problem effectively. The proposed CNN has six layers followed by a densely connected artificial neural network which is trained over the preprocessed data. The model efficiently employed the kernels and activation functions toward improving the classification performance. The proposed model effectively classifies plant leaves into disease or not disease, and further, disease plant leaf is also classified with respect to the disease very quickly due to the optimal model parameters. The experimental results demonstrate the accurate and efficient performance of the proposed method. The performance of the proposed model is evaluated on publicly available PlantVillage dataset and the quantitative results are obtained as 0.95, 0.94, and 0.94 in terms of precision, recall, and F1-score, respectively.