Mango Dataset: A Comprehensive Resource for Agricultural Research and Disease Detection

The Mango Leaf Disease Detection Dataset is expected to become a benchmark resource for researchers, practitioners, and stakeholders involved in agriculture, machine learning, and computer vision to detect and classify diseases affecting the mango plant. This dataset is intended to contribute to furthering agricultural technology by improving crop protection strategies and enabling effective machine-learning model development for disease diagnosis.

It includes high-resolution images of mango leaves that were collected from August 15 to August 29, 2023, from the mango garden of Supu Ashulia, Bangladesh. In all, there are 1,319 original images at a resolution of 1000×1000 pixels showing various conditions of the mango leaves in natural environmental conditions. The classes of the dataset are divided into five main classes, namely Anthracnose, Die Black, Gall Midge, Powdery Mildew, and Healthy. These categories represent different diseased statuses and healthy conditions of the leaves, therefore providing a very fair platform for training and testing the model.

Class imbalance has been addressed, and the utility of the dataset improved through augmentation techniques. Each class is expanded to 1,000 images, totaling 5,000 augmented images. Flipping, rotation, scaling, and other augmentation techniques have ensured increased diversity and robustness in the dataset. The augmented images are supposed to provide better generalization by the model and an accuracy increase in detecting and classifying mango leaf diseases across a wide set of conditions.

Folder Structure:
1. Original Mango Leaf Disease Dataset:
Number of datasets: 1,319
Data format: .jpg

2. Augmented Mango Leaf Disease Dataset:
Number of datasets: 5,000
Data format: .jpg