# Comprehensive data of 10 fruit leaf classes captured for agricultural AI applications

## TL;DR

This paper introduces a high-quality dataset of 10 fruit leaf species from Bangladesh, suitable for agricultural AI applications.

## Contribution

The dataset provides region-specific, high-resolution leaf images with standardized smartphone imaging and controlled lighting.

## Key findings

- The dataset contains 3173 images of 10 fruit species, with 300-343 images per class.
- Images were captured using high-resolution smartphones under controlled lighting for consistent quality.
- The dataset supports machine learning and computer vision applications for agricultural monitoring.

## Abstract

This corpus contains 3173 high-quality images of leaves of ten commonly found fruit species in Bangladesh, namely Lotkon (306), Lychee (312), Mango (330), Black plum (304), Custard apple (304), Guava (325), Jackfruit (311), Aegle marmelos (336), Star Fruit (343), Plum (302). It is captured with Realme 7-Pro (64 MP primary camera) and Realme 8-Pro (108 MP primary camera) smartphones at nurseries near to Daffodil International University, Bangladesh. This dataset addresses the scarcity of high-quality, region-specific agricultural image datasets in South Asia, offering a unique combination of standardized smartphones-based imaging and controlled lighting to ensure consistant high-resolution visual data compared to existiong datasets. To ensure uniform image quality, all leaf specimens were photographed in controlled lighting against a white paper background. The dataset has a fairly balanced number of photos for each class, with between 300 and 343 photos for each class which makes it suitable for machine learning applications. To capture a complementary range of visual properties (leaf shape, venation patterns, edges, surfaces, etc.), the collection contains healthy leaf samples photographed from both the top and underside angels. This collection fulfills the need for South Asian region-specific datasets for agricultural images and could be utilized to develop computer vision applications, automated crops recognition systems, and agricultural monitoring software. The complete dataset is public and can be accessed on Mendeley Data repository and is hierarchically structured with separate directories for all the fruit species.