Model Comparison Study for Microscopic Symptom Classification of Plant Diseases

This study is an important component of an innovative project that integrates agriculture with artificial intelligence, aimed at detecting difficult diseases that are challenging for humans to identify. Leveraging the power of artificial intelligence, our work focuses on pinpointing imperceptible diseases. The performance of the classification model depends on the informativeness of the feature map. Therefore, we compare various models specialized for this purpose. Given the characteristics of plant diseases, many complexities make it challenging to distinguish between different illnesses. In this study, we reaffirm the exceptional performance of the Residual Attention Network (RAN) consisting of residual blocks and attention modules. We have previously reviewed a paper on the combined Generative Adversarial Network (GAN) of residuals and attention, which has proven advantageous in exploring the inherent complexities of these challenges. The primary goal of our study is to utilize artificial intelligence to identify the best-performing disease classification model. When the classification problem was executed with RAN, significant improvements in all indicators were achieved.