Real-time mandarin plant species classification using deep learning approach

Abstract

In this study, the research into the immediate classification of Mandarin plant species by applying a deep learning method. The more and more people want to find the best way to identify the plant species, especially in agriculture and ecology, our research aims at using Convolutional Neural Networks (CNNs) to classify the plants based on the leaves. The study, called "Real-Time Mandarin Plant Species Classification Using Deep Learning Approach," is based on the creation of a dataset that contains images of Mandarin plant leaves. We use the latest CNN architectures, such as MobileNetV2, DenseNet201, and InceptionV3, to train the classification models which can identify the Mandarin plant species in real-time scenarios with the highest accuracy possible. Our approach consists of image pre-processing, dataset augmentation to increase model accuracy, and training of the CNN architectures using transfer learning techniques. After that we compare the performance of each model using the accuracy, precision, recall, and F1 score. We, by means of a lot of experiment and analysis, evaluate the strengths and weaknesses of all the CNN architectures in the classification of the Mandarin plant species based on the leaf images. In general, this research increases the level of plant species classification methods, thus, giving the analysis of the CNN architectures performance in the field of Mandarin plant species classification. The results show that it is important to use the deep learning methods for the accurate and efficient botanical classification tasks and thus the agricultural production, the conservation of the environment and the research of biodiversity will be improved.