Multi-Plant Disease Detection and Classification Using Multi-Scale Dynamic Spatial Gating - Squeeze and Excitation (MDSG-SE)

Abstract

The increasing influence of plant illnesses on worldwide agricultural yield demands sophisticated technologies for prompt identification and categorization. Multi-Plant Disease Detection and Classification using Multi-Scale Dynamic Spatial Gating - Squeeze and Excitation (MDSG-SE) is a novel approach that is introduced in this study. The suggested model improves the ability to discriminate in plant imagery by utilizing multi-scale dynamic spatial restricting to capture complex patterns at various scales. By combining the Squeeze and Excitation mechanisms, MDSG-SE allows the model to adaptively recalibrate the responses of individual channel features. By improving the network's empathy to disease-specific features, this dynamic recalibration maximizes classification accuracy. The multi-scale feature enables the model to manage plant diseases with different spatial features. Using a variety of datasets, a thorough performance analysis is to demonstrate the model's resilience across various crops. According to experimental results, MDSG-SE performs better than current techniques for both recognition and categorization accuracy. The model is resilient to changes in the environment and shows outstanding responsiveness to subtle disease symptoms. The general efficacy of plant disease identification systems is synergistically enhanced by the combination of multi-scale fluid spatial gating, squeeze, and excitation in MDSG-SE.