Enhancing Coconut Tree Diseases Detection with Efficient-NASNet using xAI

The coconut industry possesses considerable economic importance, especially in tropical areas. The global market for coconut products was valued at ${\$}$13 billion in 2019 and is expected to reach ${\$}$31 billion by 2024, encouraged by rising consumer demand for coconut-based products. Coconut production faces difficult challenges due to different diseases such as Bud Root Dropping, Bud Rot, Grey Leaf Spot, and Leaf Stem Bleeding, causing economic losses and increased production costs. This study proposes a novel approach for detecting coconut tree diseases by employing a parallel convolutional neural network (PCNN) model using Efficient-NASNet and Explainable AI. The model classifies diseased coconut trees, specifically focusing on Bud Root Dropping, Bud Rot, Gray Leaf Spot, Leaf Stem Bleeding, using a dataset of 5,798 images preprocessed with techniques such as resizing, color inversion, data augmentation, and outlier handling. Achieving an accuracy of 99.37%, the model’s performance is further validated through precision, recall, and F1-scores of 100%, 99.68%, and 99.84%, respectively. The model’s decision-making process was explained using explainable AI techniques like Grad-CAM and LIME, which gave important insights into the key features affecting the model’s classifications. This study aims to contribute to sustainable agricultural practices by offering precise disease detection solutions, reducing pesticide overuse, and supporting the coconut industry’s long-term sustainability.