Multi-Level Classifier for Medicinal Plant Type and Health Prediction

Plants fulfill the food, shelter, fuel and other life related requirement from ancient time. Human is continuously performing a lot of efforts to overcome the disease related to plant decays. To cure any disease it is necessary to identify whether a leaf is healthy or diseased. This work helps to identify two important attributes from an plant image i.e. plant leaf type and to predict leaf is either healthy or unhealthy. This combined phenomena is achieved  by multilevel classification. Well accepted publically available SMVU Katra dataset is used. Multi-class classification paradigm is used for classification on both levels i.e. plant leaf type and plant leaf health. A brief discussion with six machine learning classifiers are presented. These six classifiers are Support Vector Machine(SVM), Native Bayes(NB), Logistic Regression(LR), Extreme Gradient Boosting (XGBoost), k-Nearest Neighbours (KNN) and Random Forest(RF).Random Forest classifier reports the best result with accuracy 86.32% and F1-Score 86.22 %. This work is presents a state of art framework for leaf type and leaf health detection.