An Automated System Based on Informational Value of Images for the Purpose of Classifying Plant Diseases

The diagnosis of plant illnesses in conjunction with the use of a deep learning algorithm may provide a cost-effective preventative impact, and this method has seen widespread application. Nevertheless, the area of intelligent plant disease detection still struggles with issues such as inadequate data and erroneous categorization. In the current study, an image data contribution assessment approach that is based on the assessment of inter-class resemblance is proposed with the intention of resolving the issues that have been raised. When used in conjunction with an active learning Image selection approach, this method may provide useful guidance for the creation and annotation of database of plant diseases. This can increase the recognition impact while simultaneously lowering the associated costs. Both the inter-classes similarity (ICS) assessment module and the image data contribution Evaluation module are included into the suggested technique. As a result, the categorization of plant diseases will benefit from the additional information they supply. Experiments were performed using a dataset including fine-grained classifications of tomato illnesses so that it could be determined whether or not this strategy is successful. The results of the experiments demonstrate that this strategy is much more effective than others. This study is being done in the realm of classifying different plant diseases. Additional study is recommended for both the detection and the categorization processes.