Accurate Prediction and Classification of Corn Leaf Disease Using Adaptive Moment Estimation Optimizer in Deep Learning Networks

The identification and classification of plant diseases in the crop is an important aspect of the agriculture sector. The transfer learning approach in the deep learning model of the Alexnet architecture with the Adaptive Moment Estimation (ADAM) optimizer is used for training which combines the advantages of RMSprop and Stochastic Gradient Descent with the momentum (SGDM) algorithm. The 25-layer Alexnet model with 5 convolutional layers, 3 max-pooling layers, 2 normalization layers, 2 fully connected layers, and 1 softmax layer is considered for classifying 5300 images divided into four categories: healthy, blight, common rust, and gray leaf spot. Various hyperparameter settings such as learning rates, number of epochs, training–testing ratio, and different optimizers such as ADAM, SGDM, and RMSprop are tuned to train the proposed model. The experimental results for various learning rates are presented. A comparison with other existing approaches revealed that our proposed approach produced an accuracy of 99.43% for an optimal learning rate of 0.0001.This proposed work will find application in many agricultural sectors for implementation of the automated systems like robotic pesticide sprayers and drone operated systems.