A field-acquired tea leaf image dataset for disease and pest-damage classification in Bangladesh
- 1Corresponding author.
- 2Data Science Lab, Department of Software Engineering, Daffodil International University, Daffodil Smart City, Birulia, Dhaka, 1216, Bangladesh
Abstract
This data article presents a field-acquired collection of 2500 tea-leaf photographs from four tea-growing locations in Chunarughat, Habiganj, Bangladesh, collected between April 07 and June 23, 2025. The repository contains four image-level categories: Blight (Brown and Gray Blight combined; 555 images), Red Rust (620 images), Helopeltis damage (675 images), and finally Healthy Leaf (650 images). Images were captured under naturally varying illumination, leaf orientation, and background conditions and were retained at their available photographic dimensions. The deposited JPEG images retain their original spatial characteristics, with seven distinct resolutions ranging from 1200 × 1600 to 4608 × 3456 pixels (2411 portrait and 89 landscape images; no square images). The collection was acquired using multiple smartphones and preserves natural field backgrounds rather than standardized laboratory scenes. Manual visual quality screening was used to exclude substantially blurred, unclear, and obvious repeated captures, while no artificial augmentation was applied to the deposited files. The principal contribution of the dataset is the release and documentation of the complete 2500-image field repository for disease- and pest-damage classification; an approximately balanced 2000-image subset was previously used for downstream deep-learning experiments. The complete dataset is openly available through Mendeley Data under a CC BY 4.0 license.
Related
- Handwriting-based Alzheimer’s disease classification on the DARWIN dataset: Repeated nested validation, task-level reduction, and cross-validated explainability
- Citrus leaf pathology multi-class image dataset and a hybrid multi-branch attention fusion network for lemon leaf disease classification
- Low levels of Fall Armyworm (Spodoptera frugiperda J.E. Smith) damage in Bangladesh are linked to farmers’ integrated pest management practices and weather conditions
- FieldMaizeLeaf-8: A real-field early vegetative maize leaf stress image dataset from Bangladesh
- An annotated cucumber leaf image dataset for detection of visible damage and abnormal color symptoms in complex cultivation environments
BibTeX
@article{chowdhury2026fieldacquired,
title = {{A field-acquired tea leaf image dataset for disease and pest-damage classification in Bangladesh}},
author = {Sakibul Hasan Chowdhury and Md Shohel Arman},
journal = {Data in Brief},
year = {2026},
month = aug,
doi = {10.1016/j.dib.2026.113356},
url = {https://doi.org/10.1016/j.dib.2026.113356},
note = {CC BY},
}
RIS
TY - JOUR
TI - A field-acquired tea leaf image dataset for disease and pest-damage classification in Bangladesh
AU - Sakibul Hasan Chowdhury
AU - Md Shohel Arman
PY - 2026
DA - 2026/08/01
JO - Data in Brief
DO - 10.1016/j.dib.2026.113356
UR - https://doi.org/10.1016/j.dib.2026.113356
AB - This data article presents a field-acquired collection of 2500 tea-leaf photographs from four tea-growing locations in Chunarughat, Habiganj, Bangladesh, collected between April 07 and June 23, 2025. The repository contains four image-level categories: Blight (Brown and Gray Blight combined; 555 images), Red Rust (620 images), Helopeltis damage (675 images), and finally Healthy Leaf (650 images). Images were captured under naturally varying illumination, leaf orientation, and background conditions and were retained at their available photographic dimensions. The deposited JPEG images retain their original spatial characteristics, with seven distinct resolutions ranging from 1200 × 1600 to 4608 × 3456 pixels (2411 portrait and 89 landscape images; no square images). The collection was acquired using multiple smartphones and preserves natural field backgrounds rather than standardized laboratory scenes. Manual visual quality screening was used to exclude substantially blurred, unclear, and obvious repeated captures, while no artificial augmentation was applied to the deposited files. The principal contribution of the dataset is the release and documentation of the complete 2500-image field repository for disease- and pest-damage classification; an approximately balanced 2000-image subset was previously used for downstream deep-learning experiments. The complete dataset is openly available through Mendeley Data under a CC BY 4.0 license.
SN - 2352-3409
ER -