Comprehensive Lemon Leaf Disease Dataset for Advanced Detection and Sustainable Agriculture

The diseases of the lemon leaf are one of the major obstacles to its culture, leading to reduced yields and economic losses in agricultural areas like Bangladesh. The early detection and proper classification of these diseases will help in the efficient management of diseases and the protection of the crops. The manual detection of such diseases is laborious and hence might not be perfect; there is a dire need for automatic systems based on modern computer vision techniques.

In the backdrop of this context, we introduce a complete Lemon Leaf Disease Detection Dataset to support the development of models that will automatically detect diseases. We provide high-resolution images of healthy and diseased lemon leaves from various prevalent disease conditions in Bangladesh. Our dataset consists of a total of nine classes of images which include anthracnose, bacterial blight, citrus canker, curl virus, deficiency leaf, dry leaf, healthy leaf, sooty mould, and spider mites. These classes were prepared with the help of agricultural experts to make the labeling and classification accurate.

A total number of 1,354 raw images were captured manually from Melandaha, Jamalpur with the help of an agriculture expert from July to September 2024. The images were captured using an iPhone 14 Plus with a resolution of 800×800pixels to ensure quality visual data for training and evaluation. Further enhancement of the dataset was made by generating 9,000 augmented images using different augmentation techniques: flipping, rotation, zoom, shear, and brightness. This, in turn, makes the dataset even more diverse, which enhances the performance of deep learning models across different scenarios.

The Lemon Leaf Disease Detection Dataset will help researchers and practitioners develop and improve the performance of machine learning models for precision agriculture. A diversified and well-structured dataset like this is bound to help one construct an algorithm of high performance for early disease detection that could support sustainable agricultural practices in improving crop yield outcomes.

1. Original Dataset:
Number of datasets: 1354
Data format: .jpg

2. Augmented Dataset:
Number of datasets: 9000
Data format: .jpg