Artificial Intelligence: For the Recognition of Plant Species in the Malvaceae Family

Abstract

This study aimed to classify leaves of selected plants from the Malvaceae family using convolutional neural networks and deep learning techniques. We utilized 170 images from three species within this family for training. By exploiting CNNs with deep learning, we enhanced efficiency and classification accuracy, significantly reducing training time by using pre-trained models. The study was conducted with Python and various libraries, including NumPy, scikit-learn, and OpenCV. Our CNN model successfully distinguished between the different leaf images, achieving an accuracy of 71%. This approach highlights the benefits of transfer learning, particularly with limited data, and emphasizes the importance of leaf characteristics—such as shape, texture, and venation—in accurate classification. The results showcase the advancements deep learning has made in visual recognition tasks, suggesting future work could expand the dataset, explore other advanced architectures, and apply data augmentation techniques to improve accuracy. Ultimately, this research contributes valuable insights for ecological studies, agriculture, and conservation efforts by aiding in the identification and monitoring of plant species.