A comprehensive lemongrass (Cymbopogon citratus) leaf dataset for agricultural research and disease prevention

Abstract

This article introduces a dataset of 10,042 Lemongrass (Cymbopogon citratus) leaf images, captured with high quality camera of a mobile phone in real-world conditions. The dataset classifies leaves as “Dried,” “Healthy,” or “Unhealthy,” making it useful for machine learning, agriculture research, and plant health analysis. We collected the plant leaves from the Vishwakarma University Pune herbal garden and the captured the images in diverse backgrounds, angles, and lighting conditions. The images underwent pre-processing, involving batch image resizing through FastStone Photo Resizer and subsequent operations for compatibility with pre-trained models using the ‘preprocess_input’ function in the Keras library. The significance of the Lemongrass Leaves Dataset was demonstrated through experiments using well-known pre-trained models, such as InceptionV3, Xception, and MobileNetV2, showcasing its potential to enhance machine learning model accuracy in Lemongrass leaf identification and disease detection. Our goal is to aid researchers, farmers, and enthusiasts in improving Lemongrass cultivation and disease prevention. Researchers can use this dataset to train machine learning models for leaf condition classification, while farmers can monitor their crop's health. Its authenticity and size make it valuable for projects enhancing Lemongrass cultivation, boosting crop yield, and preventing diseases. This dataset is a significant step toward sustainable agriculture and plant health management.