BDMaizeSeeds: A Curated RGB Image Dataset of Commercial Maize Seed Varieties in Bangladesh

Published: 3 August 2026| Version 2 | DOI: 10.17632/nhfvmfpm6f.2
Contributors:
,

Description

BDMaizeSeeds is a publicly available RGB image dataset containing 7,800 standardized images of ten commercially cultivated maize seed varieties collected from authenticated sources across Bangladesh. Images were acquired under controlled imaging conditions using a fixed camera setup and are provided as standardized (512×512) and preprocessed (224×224) versions. The dataset also includes an augmented training subset containing 38,220 images generated exclusively from the training partition to support reproducible deep learning research and comply with project data-sharing requirements. The dataset is intended for image classification, fine-grained visual recognition, transfer learning, visual similarity analysis, explainable artificial intelligence (XAI), and other agricultural computer vision applications.

Files

Steps to reproduce

Download the standardized dataset (512×512) or the preprocessed dataset (224×224). Use the predefined training partition to reproduce the preprocessing, augmentation, and deep learning experiments described in the accompanying publication. The augmented dataset was generated exclusively from the training subset and should not be used for validation or test evaluation.

Institutions

Categories

Computer Science, Artificial Intelligence, Computer Vision, Machine Learning

Licence