To speed up disease detection and enhance productivity, new technologies have been implemented in modern farming. Accurate and early identification of plant disease diagnosis enhances plant health and the productivity of food crops. Traditional machine learning- and deep learning-based image classification models are implemented to identify plant diseases. However, these techniques may face difficulties in handling huge datasets, which contain several types of plant diseases with environmental variables. To conquer the existing challenges, this work designed an effective leaf disease classification model using deep learning networks. The images of plant leaves are initially gathered from standard datasets. Further, the gathered images are passed to the Weighted Ensemble Vision Transformer-based Convolution Network (WEViTCNet) classification model. Here, the Vision Transformer (ViT) features, Visual Geometry Group (VGG16) features, DenseNet and Xception features are extracted to identify the key features from the data. Moreover, the weighted feature fusion is performed to focus on more relevant features, where weight optimization takes place using the Iteration-aided Red-tailed Hawk Algorithm (IRHA). Finally, the weighted fused features are passed to the remaining layers of the convolution network to obtain the classified outcome. All techniques used in this network are trained on the same dataset. Then, the effectiveness of the designed plant disease segmentation and classification model is contrasted with various baseline models. The results show that the proposed model attains a better accuracy rate of 94.8%. Hence, the proposed model achieves the desired result and superior execution over plant disease classification than other conventional techniques.
Priyatharsini et al. (Sat,) studied this question.