A Robust Deep Learning Model for the Detection of Diseases in Cocos Nucifera

The detection of diseases in Cocos Nucifera (coconut trees) is essential for maintaining healthy crops and optimizing agricultural practices. In this study, we propose a deep learning-based approach for disease detection, comparing several existing models and introducing novel hybrid models. The existing models include EfficientNet, VGGNet, MobileNet, and ResNet, each known for their performance in image classification tasks. However, these models often struggle to capture intricate temporal and spatial patterns crucial for accurate disease identification in agricultural settings.

To address these limitations, we propose two novel hybrid models: the CNN-BiLSTM Hybrid and EEGNet. The CNN-BiLSTM Hybrid model combines the power of Convolutional Neural Networks (CNNs) for feature extraction and Bidirectional Long Short-Term Memory (BiLSTM) networks for capturing sequential dependencies, enabling the model to better handle variations over time in disease manifestations. EEGNet, originally designed for brainwave signal analysis, is adapted here to leverage its efficient feature extraction and classification capabilities.

Additionally, we explore a hybrid model combining MobileNet with Support Vector Machines (SVM) to enhance classification accuracy, balancing computational efficiency and performance.

The model classifies the following disease categories: WCLWD_Yellowing, CCI_Caterpillars, Healthy_Leaves, CCI_Leaflets, WCLWD_Flaccidity, and WCLWD_DryingofLeaflets. These classes represent various common diseases and conditions affecting coconut trees. The results demonstrate the superiority of the proposed hybrid models over the existing ones, with improved accuracy, robustness, and real-time applicability in agricultural disease monitoring.

This research contributes to the development of automated systems for early disease detection in coconut plantations, facilitating timely intervention and minimizing crop loss.

Keywords: Cocos Nucifera, Deep Learning, Disease Detection, CNN, BiLSTM, EEGNet, Hybrid Models, MobileNet, SVM, Agricultural Monitoring, Image Classification.