IMPROVED YOLOV8 ALGORITHM FOR FRUIT AND VEGETABLE DISEASE DETECTION

To address the issues of low accuracy and missed or false detection technology for fruits and vegetables, an improved YOLOv8 algorithm named YOLOv8-GFPN is proposed. The GFPN network was used to replace the original YOLOv8 Neck network to enhance the model’s feature extraction ability. C2f-fast-EMA was employed to replace the original C2f module, reducing model parameters and computational complexity. Wise-IoU was introduced to replace the original CIoU loss function, improving overall model performance. To verify the advancement of improved YOLOv8-GFPN algorithm, the improved algorithm was compared with original YOLOv8 on fruit and vegetable disease dataset. mAP@0.5 increased by 2.9 percentage points, mAP@0.5:0.95 increased by 5.1 percentage points, GFLOPs reduced by 17%. Experimental results show that the improved algorithm is more suitable for fruit and vegetable disease recognition.