Mango Variety and Grading Dataset

Published: 6 July 2021| Version 1 | DOI: 10.17632/5mc3s86982.1


This dataset contains images of eight varieties of Pakistani mangoes. An experiment is performed on the proposed dataset for automated classification and grading of harvested mangoes to facilitate farmers in delivering high-quality mangoes on time for export, and a high accuracy is achieved using Convolutional Neural Network. Researchers and students can use this dataset to develop, test and evaluate different computer vision algorithms to contribute the improving agriculture sector. The provided dataset can be consider as a benchmark for testing and comparing the performance of different state-of-the-arts. We would like to thank the Haji Ghulam Muhammad Mangana mango farm (Registered) Multan Pakistan for database collection and (MRI) Mango Research Institute, Multan for guidance about the standardized grading of mango.


Steps to reproduce

Mangoes were collected from Haji Ghulam Muhammad Mangana mango orchard in Nawabpur (Multan) Pakistan for database collection. The best time for image acquisition is from around twenty days after starting the stone-hardening stage of mango development to one week before the start of mango harvesting at the end of June. Images have been captured using Nikon 7000 camera between 9 AM to 12 AM in the natural light intensity. The image capturing task has been carried out in MNS-University of Agriculture Multan Pakistan from June to August 2020. The camera was fixed on a tripod to control its movement in order to avoid blurred and fuzziness of the mango image, and white background was used for easy and accurate segmentation.


Muhammad Nawaz Shareef University of Agriculture


Artificial Intelligence, Mango, Database