Large-Scale Lemon Leaf Disease and Pest Image Dataset from Bangladesh
Description
This dataset contains citrus leaf images captured under real farm conditions across four major lemon-producing regions in Bangladesh. The images represent a wide range of citrus leaf diseases and pest attacks, including both common infections and nutrient-deficiency symptoms. Each image has been manually labeled and classified into specific disease or pest categories by agricultural domain experts. The dataset is intended for research purposes in the fields of artificial intelligence, computer vision, and plant pathology ā especially for training deep learning-based early detection systems for citrus diseases and pests. Algal Leaf Spot (875 Images) Anthracnose (1079 Images) Bacterial Blight (244 Images) Black Spot (837 Images) Citrus Canker (1789 Images) Citrus Hindu Mite (544 Images) Citrus Leafminer (900 Images) Citrus Pest (545 Images) Citrus Scab (333 Images) Curl Leaf (1853 Images) Dry Leaf (390 Images) Greening (2340 Images) Healthy (1644 Images) Lemon Sooty Mold (1597 Images) Melanose (376 Images) Spider Mites (250 Images) Swallowtail Larval Herbivory (Deficiency) (1343 Images) Yellow Spot (670 Images) š Total Images: 17,609 Image Properties: Format: JPG & PNG Resolution: 1000 Ć 1000 pixels Color space: RGB Capture device: Smartphone cameras Background: Variable ā Real farm leaves, different angles, lighting, and white backgrounds Image Collection Locations Citrus Research Station ā Jaintapur, Sylhet Lemon Garden Resort ā Sreemangal, Moulvibazar Thai Lemon Orchard ā Jaldhaka, Nilphamari Seedless Lemon Farm ā Boalmari, Faridpur
Files
Steps to reproduce
The dataset was developed through a structured pipeline consisting of field data collection, expert validation, and systematic image processing. Lemon leaves were photographed directly in real farm environments across four agricultural locations in Bangladesh. Diseased, pest-infected, nutrient-deficient, and healthy leaves were first identified by field experts. Images were captured using smartphone cameras under diverse lighting conditions, angles, and backgrounds to preserve the natural variability of farm scenarios. After collection, low-quality or duplicate images were removed, and all remaining images were normalized to a resolution of 1000 Ć 1000 pixels in RGB format (JPG or PNG). Agricultural specialists then manually reviewed each image to classify it into one of the 18 categories, ensuring high-quality, consensus labeling. The images were sorted into class-wise folders and label information was recorded for dataset accessibility. Quality checks were performed through random validation to eliminate any incorrect or ambiguous labels. The same workflow can be reproduced using a modern smartphone camera, basic image preprocessing tools (e.g., Photoshop, GIMP, or Python libraries such as OpenCV/PIL), and expert guidance from citrus disease specialists for accurate category assignment.