ACCDW-YOLO: an effective detection method for small-sized pests and diseases in navel oranges

The identification of pests and diseases in navel orange trees is vital for safeguarding orchard production quality and yield. A termed ACCDW-YOLO model is proposed based on the YOLOv7 model to achieve low-cost and high-precision detection for Navel Orange pests and diseases with small sizes in natural environments. The model optimizes the YOLOv7 architecture by integrating the ACmix mechanism. This integration significantly elevates the model's learning and inference capabilities. The incorporation of the DCNv3 module into the design creates the DCNv3-E-ELAN module, which enhances the model’s proficiency in detecting objects of varying sizes. The Wise Intersection over Union version 3 (WIoUv3) loss function was used to reduce the competitiveness of high-quality anchor boxes, reduce the harmful gradients generated by low-quality samples, and improve the overall performance of the model. The results of the tests using IP102 dataset and real orchards dataset showed that the ACCDW-YOLO model showed a significant improvement in small-target detection, with an mAP0.5 of 88.8% and 69.7%, respectively, which is superior to other state-of-the-art models. This model provides a new benchmark for pest and disease identification in navel orange cultivation.