Annotated Mahua Leaf Dataset for Disease Detection and Precision Agriculture

Published: 21 July 2026| Version 1 | DOI: 10.17632/6xjn6p73h7.1
Contributors:
Monica Sankat,

Description

The Annotated Mahua Leaf Image Dataset is a curated collection of high-resolution images of Mahua (Madhuca longifolia syn. Madhuca indica) plants captured under natural field conditions in Madhya Pradesh, India. The dataset was developed as part of the Madhya Pradesh Council of Science and Technology (MPCST), Government of Madhya Pradesh, funded research project titled "Development of an Autonomous Robotic Device for Detection, Precision Spraying, and Disease Control in Mahua Cultivation." This dataset extends our previously published raw Mahua leaf image dataset by providing manually annotated bounding-box labels for healthy and unhealthy Mahua leaves. The annotations enable the development, training, validation, and benchmarking of artificial intelligence and computer vision models for automated disease detection and object detection in precision agriculture. The images were acquired under diverse real-field conditions, including variations in illumination, background, leaf orientation, growth stages, and environmental conditions, thereby reflecting realistic agricultural scenarios. Each image has been carefully annotated using bounding boxes to identify healthy and diseased leaf regions, providing high-quality ground truth for supervised learning applications. To improve reproducibility and support robust model development, the dataset has been standardized through image preprocessing, quality verification, annotation validation, and a leakage-free train, validation, and test split. It is suitable for a wide range of applications, including object detection, disease diagnosis, computer vision, deep learning, federated learning, edge artificial intelligence, drone-assisted crop monitoring, autonomous robotic systems, and precision agriculture. The dataset is intended to serve as a benchmark resource for researchers, academicians, and developers working on intelligent agricultural systems and AI-enabled plant health monitoring.

Files

Steps to reproduce

1.Download the dataset from the Mendeley Data repository. 2.Extract the compressed archive while preserving the original folder structure. 3.The dataset contains field-acquired Mahua (Madhuca longifolia) images along with manually annotated bounding-box labels for healthy and unhealthy leaves. 4.Load the images and corresponding annotation files using any compatible object detection framework (e.g., Ultralytics YOLO, Detectron2, MMDetection, or TensorFlow Object Detection API). 5.Use the provided train, validation, and test splits (or create custom splits if required) for model development and evaluation. Apply standard image preprocessing techniques such as resizing and normalization according to the requirements of the selected deep learning model. 6.Train and evaluate object detection models using the provided annotations to detect healthy and unhealthy Mahua leaves. 7.The dataset can be used for benchmarking algorithms in object detection, plant disease detection, precision agriculture, federated learning, drone-based crop monitoring, and autonomous robotic systems.

Institutions

Categories

Agricultural Plant, Federated Learning, Agriculture

Funders

Licence