Coconut Disease Prediction Using Neural Attention Mechanism With Adamax Optimizer Based Convolutional Neural Network

Abstract

Accurate coconut disease prediction systems become most important in developing effective disease mitigation strategies, encouraging cost-effective crop protection for small-scale farming. The farmers utilize only the conventional methods like laboratory view and experts for the identification of diseases in coconut. However, these techniques are inadequate for an effective and timely detection of diseases in the coconut. Hence, this research proposes the Neural Attention Mechanism with AdaMax Optimizer based Convolutional Neural Network (NAM-AMO-CNN) approach for prediction of coconut disease. This article is comprises of different levels: Initially, the dataset is collected from the real-world of various images of coconut diseases. Next, pre-processing is conducted through data augmentation and noise removal through Quaternion Non-local Means Denoising Algorithm (QNLM) approach. Then, segmentation is performed through DeepLabV3+ and Feature extraction is employed by Residual DenseNet. Finally, prediction is performed through NAM-AMO-CNN approach, which allows the classifier to concentrate on critical disease-infected regions, leading to improved accuracy. Experimental results demonstrates that the proposed NAM-AMO-CNN approach attains the superior accuracy of 98.43% as compared to the existing methods like Artificial Intelligence Enabled Coconut Tree Disease Detection and Classification (AIE-CTDDC) and Enhanced Visual Geometry Group (EVGG16).