Plant Disease Recognition Using Residual Convolutional Enlightened Swin Transformer Networks

Abstract

Adoption of cutting-edge technology such as the Internet of Things (IoT) as well as Deep Learning ideas in smart agricultural system has expanded dramatically in recent years in order to ensure humanity's long-term viability. It is critical to distinguish the plant photos captured for monitoring and, as a result, disease and pest prevention. Currently, deep learning methods are employed for an efficient recognition of plant disorders. However, categorization and recognition of plant diseases is a difficult task due to the presence of low-intensity data in the image, the presence of noise in the sample, and differences in the position, chrominance, and colour similarities among both healthy and diseased plants. plant images. Hence, the overarching objective is to design the custom and efficient deep learning model that can yield an accurate identification of plant diseases under the above mentioned real time environment. In this research article, novel ensemble of residual convolutional networks and Swin transformers has been proposed. Swin Transformers (ST) is hierarchal architecture which brings the greater efficiency and flexibility at various scales and has linear computational complexity with respect to the image size. Therefore, combining the Swin Transformers and powerful Residual networks will extract the numerous deep key point features that aids for the better classification and recognition of plant diseases under different real time circumstances. PlantVillage Kaggle datasets are used for extended experimentation, and various performance measures including as accuracy, precision, recall, specificity, and F1-score are computed and analysed. To prove the superiority of the proposed model, existing architecture such as FCN-8 s, CED-Net, SegNet, DeepLabv3, Densenets and Centernets. The experimental findings show that the suggested model outperformed the other existing techniques in terms of accuracy, precision, recall, and F1-score. Also comprehensive analysis shows that the proposed model is more capable and reliable in identifying diseases in plants under real time circumstances.