A comprehensive dataset of agarwood tree ( Aquilaria Malaccensis ) leaf images for disease analysis in Brunei Darussalam.

Abstract

The visual diagnosis based on foliar traits remains a cornerstone technique for the early identification of biotic stress, for instance, disease and pest infestations, in many economically valuable crops, including Aquilaria Malaccensis (agarwood). As a species of immense commercial and ecological significance, Aquilaria Malaccensis is particularly vulnerable to a range of pathogens and insect threats that can severely compromise resin production and tree viability. With the increasing integration of disruptive sustainable agricultural technologies, such as artificial intelligence (AI), especially in plant phenotyping and pathology, the development of robust and generalizable AI models hinges on the availability of large-scale and high-resolution image datasets. However, the current lack of such curated datasets for agarwood poses a substantial bottleneck to progress in automated identification systems. This deficiency limits the ability of scientists, technologists, and plant health experts to leverage machine learning and computer vision techniques for timely, accurate, and scalable solutions to different stresses in agarwood disease and pest management, including nematodes, viroids, viruses, pests, phytoplasmas, bacteria, fungi, and Protozoa. This paper presents a dataset of pests and diseases affecting agarwood trees, which impact farmers. It includes a total of 5472 leaf images classified into 14 categories. These categories consist of 8 types of agarwood diseases, 5 types of pests, and 1 category of healthy leaf images, encompassing both insect-damaged and healthy leaves. The images were captured using a PowerShot G7X Mark III camera. The images were captured from three different agarwood plantation sites of Batong, Benutan, and Bukit Silat in 2024, led by the Institute for Biodiversity and Environmental Research (IBER), Universiti Brunei Darussalam, by Botanical Research Centre (UBD BRC) scientists and biologists. This dataset is particularly valuable for training and validating deep learning (DL), computer vision, and machine learning algorithms aimed at identifying agarwood diseases and pests in agarwood leaves. Offering researchers and learners a robust data resource for analyzing and improving agarwood plant health through the development of advanced computational models. The designed models are vital and hold immense practical value for farmers, equipping them with the tools that timely detect and identify diseases in their agarwood trees, empowering them to make informed decisions and potentially intensify their profits.

Plant phenotyping relevance

葉画像から病害・害虫による植物状態を識別するための大規模データセットを構築しており、データ取得と再利用可能な解析基盤が研究の中心であるため。

Code and data availability

The paper is a data descriptor for a public agarwood leaf image dataset (5472 images, 14 classes) deposited on Zenodo and Mendeley, with explicit direct URLs and DOIs provided in the Data Accessibility section. This is the paper's own phenotyping image dataset, publicly available and actionable. No separate author code