Utilizing Deep Learning for the Identification and Suitability of Medicinal Plants in Disease Treatment

Abstract

Introduction: The production of medications in the Ayurvedic, folk, and herbal medicine sectors depends on the precise identification of medicinal plants. Every plant species has special qualities that add to its healing benefits. Characteristics that help identify therapeutic plants include One of the main ways to identify a species is by its leaf shape, which can differ greatly between species. Broad, narrow, lobed, and compound leaf shapes and their arranged on the stem offers important identifying cues. The health and species type of the leaves can be determined by their color on both the upper and bottom surfaces. Species, maturity, and environmental factors can all influence color variations.

Objectives: Analyse feature vectors from both the upper and lower surfaces of a green leaf, alongside various morphological characteristics, to determine the optimal combination of features that enhances the identification accuracy and the usability.

Methods: In this study, The initial stage, known as image acquisition, is taking pictures of plant leaves, typically using a camera or a pre-defined data set. The analysis will make use of the Leaf Image Dataset. To train the model, this dataset contains pictures of both healthy and sick leaves. Pre-processing is used to standardize the format, improve image quality, and eliminate noise. Training and testing are the two sections of the dataset. To automatically extract significant information from the photos, CNN (Convolutional Neural Network) is used. It is very good in image analysis. Metrics like accuracy, precision, recall, and F1-score are checked to evaluate performance. Lastly, the health statuses of the plant and suitable cures or treatments for the illness are recommended.

Results: By employing this technique, the system displays an image of a plant leaf together with its scientific name, local name, qualities, and the ailment it treats, as well as whether the leaf is a member of the medicinal plant or not. Because of its many compelling benefits, such as strengthening feature propagation and promoting feature reuse, which boost efficiency and reduce valuation loss, the Dense Net type of Convolutional Neural Network (CNN).the model is trained using Keras.

Conclusions: The recommended method identifying herbs particularly for those who are unable to use pricey analytical equipment. This study examines various plant identification techniques and weighs the benefits and drawbacks of each.