MangoClassify-12: A high-resolution image dataset of twelve indigenous Bangladeshi mango cultivars

A high-resolution image dataset, MangoClassify-12, comprising 3900 JPEG images of twelve indigenous Bangladeshi mango cultivars, was assembled to enable automated classification. Images were captured between early May and July 10th, 2025, from three distinct production regions (Mirpur-2, Dhaka; Phulbari, Dinajpur; Rajshahi) under natural light using four smartphones. All images were reviewed by agricultural experts to exclude damaged or overripe specimens. The dataset covers twelve cultivars: Amrapali, Himsagar, Harivanga, Langra, Fazli, Gopalbhog, Ranibhog, Gobindobhog, Sundari, Banana Mango, Bari-4 and Khirsapat. Metadata are organized in a structured folder hierarchy. MangoClassify-12 is openly accessible via DOI and supports machine learning applications such as variety identification, quality assessment and mobile-based recognition. By providing raw images without predefined splits or augmentations, the dataset offers a flexible benchmark for computer vision research in agriculture.