Automatic classification of ligneous leaf diseases via hierarchical vision transformer and transfer learning

Abstract

Background: Identification of leaf diseases plays an important role in the growing process of different types of plants. Current studies focusing on the detection and categorization of leaf diseases have achieved promising outcomes. However, there is still a need to enhance the performance of leaf disease categorization for practical applications within the field of Precision Agriculture. Methods: To bridge this gap, this study presents a novel approach to classifying leaf diseases in ligneous plants by offering an improved vision transformer model. The proposed approach involves utilizing a multi-head attention module to effectively capture contextual information about the images and their classes. In addition, the multi-layer perceptron module has also been employed. To train the proposed deep model, a public dataset of leaf disease is exploited, which consists of 22 distinct kinds of images depicting ligneous leaf diseases. Furthermore, the strategy of transfer learning is employed to decrease the training duration of the proposed model. Results: The experimental findings indicate that the presented approach for classifying ligneous leaf diseases can achieve an accuracy of 85.0% above. Discussion: In summary, the proposed methodology has the potential to serve as a beneficial algorithm for automated detection of leaf diseases in ligneous plants.