Enhancing the accuracy of disease detection on tomato leaves by using the YOLOv11 model combined with the WIoU loss function
Bùi Đăng Thảnh,
Nguyen Hoang Tuyen, Phuong Thuy Ngo, Le Thanh Trung
Timely and accurate identification of tomato diseases plays a key role in food supply and agricultural productivity improvement. In this study, we propose an improved CIoU loss function for the YOLOv11 deep learning model, aimed at enhancing tomato leaf disease detection. The WioU loss function introduced into the YOLOv11 model allows adaptive adjustment of the trade-off between positioning accuracy and stability of the gradient by changing the penalty in the bounding box regression according to the difficulty of each sample. Experimental evaluations on a benchmark tomato leaf disease dataset show that the YOLOv11 model with WIoU delivers enhanced detection performance, achieving an overall accuracy of 94.8%, with gains of 1.3% in mAP@50 and 1.2% in mAP@50-95 compared to the baseline model. These outcomes demonstrate that the proposed model contributes to more efficient and accurate monitoring of crop diseases, thereby reducing the burden on farmers and facilitating the practical adoption of AI in agriculture.
Năm 2026 
