Development of a Database and a Machine Learning Model for the Rapid Identification of Certain Pinaceae Species in Algeria

Abstract

In this study, we utilized various machine learning techniques, including supervised learning, unsupervised learning, and deep learning with transfer learning, to identify ten different plant species in Algeria. Our dataset comprised a combination of common and rare species, incorporating measurements and photographic data. Using a Convolutional Neural Network model based on the VGG16 architecture with transfer learning, we achieved 96% accuracy and 88% validation accuracy. Additionally, our supervised and unsupervised learning models provided perfect clustering and prediction results with 100% accuracy. Despite limited computational resources, our comprehensive approach demonstrates the potential of machine learning in enhancing plant identification accuracy and efficiency. Furthermore, we managed to design an interface for a future application called Eco Explorer that can perform the identification tasks.