Dynamic CNN combination for Morocco Aromatic and Medicinal Plant classification

Abstract

In few last years, deep learning has itself set up as the new strategy for plant classification. Deep learning has a best performance for object recognition. In this paper, we have focused on the case of Morocco aromatic and medicinal plant (AMP) classification. Leaf is an important organ of plant, it has shown satisfying performances for plant classification and recognition In addition to leaves, we have used an others organs of AMP.ie. leaf veins and branches. We have proposed a new model combining dynamically the CNN classification result using the entropy impurity method. In the experiments, we have used VGG16, ResNet50 and Inception V3 CNN models. We have used Keras with Tensor flow backend to build and compile all neural network models. The experiments present that our model shown the higher classification accuracy.