# Smartphone image dataset for machine learning-based monitoring and analysis of mango growth stages

## TL;DR

This paper introduces a standardized smartphone image dataset for monitoring mango growth stages using machine learning, specifically tailored for Bangladesh but applicable globally.

## Contribution

The paper presents a publicly accessible, annotated mango growth dataset from Bangladesh to enable machine learning applications in agriculture.

## Key findings

- The dataset includes 2004 images categorized into four mango growth stages.
- It was collected in Bangladesh but represents global mango development patterns.
- The dataset is organized with annotations to support accurate model training.

## Abstract

Machine learning and artificial intelligence have gained widespread popularity across various sectors in Bangladesh, with the notable exception of the agriculture industry. While wealthier nations have extensively adopted machine learning and deep learning techniques in agriculture, Bangladesh's agricultural sector has been slower to follow suit. A key factor in the success of any machine learning model is the availability of high-quality datasets. However, practitioners in Bangladesh's mango industry face challenges in leveraging these advanced computational methods due to the lack of standardized and publicly accessible datasets. A well-structured dataset is essential for developing accurate models and reducing misclassification in real-world applications. To address this gap, we have developed a standardized image dataset capturing different stages of mango growth. The dataset, collected between April and June at an orchard on the East West University campus in Bangladesh, consists of 2004 images, each annotated and categorized into four distinct growth stages: early-fruit, premature, mature, and ripe. Although the dataset was created using mangoes from Bangladesh, the growth stages documented are representative of mango development globally, making this dataset applicable to mango cultivation in other countries. The dataset is organized into four folders, each containing both images and corresponding annotation files. We anticipate that this dataset will serve as a valuable resource for researchers and practitioners working in the field of automated agriculture, facilitating the development of machine learning models for monitoring and analyzing mango growth stages.