Towards Large-Scale Automated Insect Pest Classification: A YOLO11-Based Approach

Abstract

Insects are one of the main influences on plants. Early detection of plant pests and accurate identification of which insect species is causing the damage will be very useful for pest control, minimizing damage in agricultural production, improving the quality and yield of agricultural products. This study takes advantage of recent advances in deep learning, employing YOLOv11 variant models to classify insect according to their species. The datasets used in this study are the IP102 public insect dataset and PEST204 an enhanced of IP102 with additional species and filtering out error images. The YOLO11 variants achieved high performance across all classes. YOLO11x-cls model achieves the highest top-1 accuracy with the shortest preprocessing time on both datasets. While YOLO11s-cls shows to be a reasonable candidate for deployment on mobile devices for real-world insect image classification tasks.