MangoLeafVarietyBD: An Image Dataset for Multi-Class Mango Variety Classification in Bangladesh

Published: 13 October 2025| Version 1 | DOI: 10.17632/hb3kvgfcvm.1
Contributors:
,
,

Description

MangoLeafVarietyBD Dataset is a carefully selected image set on the classification of mango varieties containing 2,744 raw mango-leaf shots (which were augmented to 19,208 images) of eight mango varieties: Amrapali, Banana, Bandigori, Himsagar, Surjapuri, Harivanga, Kacha Mitha, and Brunei King. The consumer mobile phones were used to capture images in natural lighting in Thakurgaon District and Daffodil International University (Savar, Dhaka) and processed with background removal and resized to 244x244px. The data is divided into class-specific folders that are compatible with pyTorch/TensorFlow ImageFolder and to facilitate transparency and reuse, the data comes with reproducible Google Colab notebooks to preprocess, augment, and train examples. It is aimed at sustaining improvement and benchmarking of working, field-deployable mango sorting systems of low-resource agrifood environments.

Files

Institutions

  • Daffodil International University

Categories

Agricultural Science, Artificial Intelligence, Computer Vision, Machine Learning, Image Classification, Plant Pathology, Convolutional Neural Network, Deep Learning, Transfer Learning

Licence