DEEP LEARNING MODEL FOR AUTOMATED DETECTION AND CLASSIFICATION OF COCONUT TREES ALIMENTS

Abstract

Coconut tree diseases present a major challenge to agricultural productivity and food security in regions dependent on coconut cultivation. Traditional diagnostic methods are labour-intensive, subjective, and often delayed, leading to significant crop loss. As agriculture increasingly moves toward automation, there is a pressing need for intelligent, data-driven solutions for early disease detection and classification. This study proposes an automated coconut leaf disease recognition model using the DenseNet-121 convolutional neural network architecture. A curated dataset of 526 images, comprising together healthy & diseased coconut greeneries, was used to prepare and evaluate the prototypical. The DenseNet-121 model was selected for its advantages in mitigating vanishing gradients, promoting feature reuse, and enhancing classification accuracy. The model was trained using transfer learning with fine-tuning, and standard pre-processing & intensification techniques were applied to progress generalizability. The proposed prototypical realized an overall classification accurateness of 99%, outperforming conventional deep learning models used in similar tasks. Precision, recall, and F1-score metrics for both healthy and diseased classes exceeded 98%, indicating high reliability and robustness. The DenseNet-121 architecture demonstrated superior performance in comparison to baseline models such as VGG16 and ResNet50, with improved convergence and feature representation. These outcome validate the helpfulness of the model in precisely sensing coconut leaf diseases.