Simple and Compound Leaf Taxonomy Embedded Machine Learning Approach for Ayurveda Plants Recognition

Ayurveda treatment is an ancient traditional method that is used to treat various diseases even in an underdevelopment society. For this, the correct identification of Ayurveda plants is mainly important, and it enhances the quality of the Ayurveda treatments. Unfortunately, many people cannot recognize these plants by seeing them at least without having an image. Sometimes even having an image, the identification can be wrong since a lot of plants have similar characteristics. Therefore, the representation of knowledge of Ayurveda plants is essential and it leads the general public to gain some knowledge that can be applied whenever they need it. This paper presents a methodology to identify Ayurveda plants widely using their leaves with images. Only leaves are available and visible all the time, this approach mainly considers about leaves and these leaves are divided into two models as the recognition of large/simple leaves model and the recognition of the small/compound leaves model. By extracting texture features and leaf shapes from around 40 different Ayurveda plants through HOG, Gabor, Hu-moments, GLCM, Canny, and Laplacian methods, then classify each with different classifiers such as Random Forest, SVM, Decision Trees, K-NN, Logistic Regression, and Gradient Boosting. The results are compared with each plant and module-wise performances with their accuracy, F1-score, precision, and recall levels.