Ayurvedic Plant Recognition Using Multi-learners Model

Plants are an important source of natural medicine as they synthesize a large number of chemical compounds to sustain their own life, against the attacks of insects, fungus, animals etc. Herbal medicines are traditionally used in many societies worldwide. The project aims for automated plant detection. The datasets used include feature dataset from Kaggle leaf Classification and feature dataset extracted from manually created leaf image dataset of Kerala plants using Histogram of Oriented Gradients(HOG) method. The model was developed after studying the performance of 7 classifiers and choosing the best 3 based on their performance metrics. The majority and weighted voting technique been used for the type prediction of plant leaves. Dimensionality reduction using Principal Component Analysis(PCA) was done without compromising accuracy and a comparative study was performed. Test results illustrate that the multi-learners approach provides better performance than the single learner's approach. Accuracy of multi-learners approximates 97–100% for Kaggle's dataset.