AgroWinterLeaf-BD: A curated field image dataset of five winter leafy vegetables from Bangladesh

Published: 10 February 2026| Version 1 | DOI: 10.17632/vzt878hwdn.1
Contributor:
Md Jahidul Alam Sagar

Description

This repository contains a field-acquired RGB image dataset of five winter leafy vegetables cultivated in agricultural plots in Cumilla, Bangladesh, captured under natural conditions between November 2025 and January 2026. Images were collected using multiple smartphone devices (e.g., iPhone 15, Redmi Note 14 Pro, vivo Y50i) to introduce sensor variability. The dataset is designed for leafy vegetable image classification and includes natural variation in illumination, leaf pose/orientation, growth stage, and background. Classes (5) 1) Purslane (Portulaca oleracea) – 231 images 2) Radish Leaves (Raphanus sativus) – 186 images 3) Red Amaranth (Amaranthus tricolor) – 337 images 4) Spinach (Spinacia oleracea) – 342 images 5) Sweet Pumpkin Leaves (Cucurbita moschata) – 248 images

Files

Steps to reproduce

The repository provides three versions of the dataset, organized into separate directories: A) Raw data (RAW_DATA/): Contains the original field-captured RGB images in .JPG format with native resolution. This version preserves all natural variations in illumination, background, viewing angle, and plant growth stage. B) Resized data (RESIZED_DATA/): Contains images resized to 224 × 224 pixels, ensuring compatibility with standard deep learning architectures while retaining the original class-wise image counts. C) Augmented data (AUGMENTED_DATA/): Contains synthetically generated training images derived from the resized dataset using common data augmentation techniques, including random horizontal flipping, rotation, color jittering, and normalization. Data augmentation was applied to balance the dataset, resulting in 342 images per class in the augmented version. Location and acquisition notes (include in description) a) Country/Region: Bangladesh b) District: Cumilla c) Acquisition setting: Real agricultural fields (in situ), natural backgrounds d) Capture period: Nov 2025 – Jan 2026 e) Image type: RGB .JPG f) Use case: Deep learning / computer vision classification of winter leafy vegetables

Institutions

Categories

Computer Vision, Vegetable, Leaf Vegetables, Deep Learning, Agriculture

Licence