A web‐based app for sunn pest species recognition (Hemiptera: Scutelleridae: Eurygaster ) using machine learning

Recent progress in machine learning, particularly in convolutional neural networks, has greatly improved insect pest identification. Eurygaster integriceps , a highly polymorphic true bug and major wheat pest across the Western Palearctic, remains difficult to distinguish from its congeners even for specialists. Misidentifications reduce monitoring accuracy and hinder cost‐effective pest control. To support reliable identification by non‐experts while maintaining high classification accuracy, we trained the MobileNetV2 model on a large dataset combining images captured under controlled conditions with various devices, including a budget smartphone and photographs taken in the wild. The trained model demonstrated high classification metrics in identifying Eurygaster species, with macro precision, recall and F1‐score values of 0.901, 0.930 and 0.912, respectively. Based on this model, we developed an open‐source, web‐based application with a microservice architecture, allowing automated species identification from user‐uploaded images. Publicly accessible at https://eurygaster.ru , this tool supports faster and more accurate field identification, helping improve pest management decisions and reduce economic losses caused by misidentification.