A lightweight deep learning model for pest and disease detection and classification in agarwood plant (Aquilaria Malaccensis Lam.)

Agarwood ( Aquilaria malaccensis , Aquilaria macrocarpa, and Aquilaria beccariana ) is a high-value crop with important economic, pharmacological, and industrial applications. However, its productivity is severely affected by pests and diseases, whose early symptoms often exhibit similar visual characteristics, making timely and accurate diagnosis challenging. Existing deep learning approaches for plant health monitoring are often computationally intensive, limiting their deployment on resource-constrained devices. This study aims to develop a lightweight and robust deep learning model for the accurate classification of novel agarwood pests and diseases. A novel convolutional neural network, termed AgarwoodNet, was designed and evaluated using a newly curated dataset comprising 5,472 agarwood leaf images collected from Batong, Benutan, and Bukit Silat in Brunei Darussalam. The dataset includes 14 classes covering major pests, diseases, and healthy leaves. The proposed model was assessed using accuracy, precision, recall, and F1-score and compared with established architectures, including GoogleNet, ResNet18, and Inception V3, among others. AgarwoodNet achieved 96.66% accuracy, 94.77% precision, 96.63% recall, and 95.46% F1-score, outperforming the benchmark models while maintaining lower computational complexity and a lightweight architecture. The results demonstrate that AgarwoodNet provides accurate, robust, and efficient classification of agarwood pests and diseases. The proposed model offers a practical solution for real-time agarwood health monitoring and early pest and disease detection, supporting smart agriculture, precision forestry, and sustainable agarwood production in resource-constrained environments.