teaLeafBD: A comprehensive image dataset to classify the diseased tea leaf to automate the leaf selection process in Bangladesh.

Abstract

Tea is an extremely popular beverage around the world due to its exquisite taste and flavor. Unfortunately, it is prone to different types of illness, which can reduce the amount of harvest along with its standard. Among these, leaf infections are a serious concern since they negatively affect the quality of tea leaves. As a consequence, tea producers often encounter a great deal of obstacles and financial losses. Keeping this in mind, a thorough dataset has been compiled, which contains 5278 images of diseased and healthy leaves. The purpose of this dataset is to improve our knowledge of how these conditions impact cultivating tea plants and tea production. These images are collected from a variety of locations and meteorological circumstances, which provide an extensive knowledge of the disease patterns unique to tea leaves. The pictures have been captured with the help of some high-quality devices from different angles and in high resolution to ensure the standard and increase the usability of the dataset. Rigorous steps were followed when preparing the dataset that would be of great help in building a precise artificial intelligence model. The dataset carefully determined and classified six tea leaf diseases: Tea algal leaf spot, Brown Blight, Gray Blight, Helopeltis, Red spider, and Green mirid bug. There is one more class in the dataset containing images of healthy leaves. These illnesses are known for their devastating impact on tea leaves. An automated disease classification system can be made utilizing deep learning techniques that will enable estate managers to take timely action to stop the spread of the disease, and this meticulously collected dataset will immensely help to train that model.

Plant phenotyping relevance

茶葉の健全・病害状態を画像で収集・分類した再利用可能なデータセットであり、植物病害状態の画像ベース表現型計測を中心とする。

Code and data availability

The paper's own teaLeafBD image dataset (5278 tea leaf images, 7 classes) is publicly deposited on Mendeley Data with an explicit DOI and direct URL, making it a paper-specific, publicly actionable asset.