DADML: Dataset for Adulteration Detection in Medicinal Leaf
Description
Ayurveda is the traditional system of medicine of India, utilizing medicinal plants for healthcare and wellness. Bioactive components present in leaves of medicinal plants possess therapeutic properties and therefore it can be used as the major ingredient in many ayurvedic formulations. Accurate identification and authentication of medicinal leaves are essential for ensuring the quality and appropriate use of herbal materials. Closely related plant species, particularly those belonging to the same family or genus, may exhibit considerable similarities in leaf morphology, making their identification difficult. Such similarities can also increase the possibility of accidental substitution or adulteration, which may affect the quality and intended use of ayurvedic preparations. Therefore, reliable identification methods are important for supporting the authentication of medicinal plants. To ensure proper identification and to prevent adulteration, we are introducing this dataset that consists of the original medicinal leaf Wrightia Tinctoria (Danthapala) along with its probable adulterant leaves such as Holarrhena Pubescens (Kudagapala), Tabernaemontana Alternifolia (Kuruttupala), Tabernaemontana Divaricata (Nandyarvattam), and young Wrightia Tinctoria (Danthapala) leaves. All these plant species are from the same family, Apocynaceae. The dataset consists of two folders: the imbalanced and balanced dataset. The imbalanced dataset consists of high-resolution 310 raw images with five classes, and each class contains 30-90 images. Balanced dataset was constructed by applying augmentations to the existing images. This dataset consists of five classes and each class consists of 90 cleaned images, resulting in a total of 450 images. This dataset has been collected from Vaidyaratnam Ayurveda Research Institute (VARI), Thrissur, Kerala, India.