Classification of Aglaonema Using Machine Learning

Recently, Aglaonema has become an ornamental plant with high economic value. Various Aglaonema variants are available in the market with various prices offered, e.g., Lipstick, Suksom, Red cochin, Widuri, and Pride of Sumatra (pos). People often need help distinguishing the variants to prevent being deceived when buying the plants. A classifier to distinguish several local Aglaonema variants with high similarities is proposed to solve this problem. A collection of images from each Aglaonema variant are used to train the classifier. Both traditional classification algorithms, such as the
$k$
-nearest neighbours and support vector machines, and modern algorithms, specifically the Inception-v3, were employed. Classifier evaluation is performed by applying 5-folds cross-validation. Experimental results show that the best level of accuracy is obtained from the Inception-v3 algorithm by achieving an accuracy rate of 83.02%. This result indicates that the classification of Aglaonema is relatively hard, but feasible to continue for improvements.