CFNet-YOLOv12: A novel deep learning framework for apple leaf disease detection

The detection of plant leaf disease is becoming increasingly important in crop health management, early disease warning, and precision agriculture. However, conventional deep learning detection methods often exhibit suboptimal performance in complex backgrounds, varying lighting conditions, leaf occlusion, and the presence of ambiguous or blurry disease features. Consequently, we propose CFNet-YOLOv12, an enhanced apple leaf disease detection method based on an improved YOLOv12 with coordinate and focal-modulated attention mechanisms. First, by leveraging the specific characteristics of apple leaf diseases, the repetition count of modules within YOLOv12 is streamlined to increase feature extraction efficiency. Second, the coordination-attention mechanism is incorporated to improve the ability of the network to extract blurry disease features. Finally, the focal-modulated attention mechanism is integrated to capture subtle features of apple leaf diseases, thereby increasing the detection accuracy. The experimental results demonstrate that the proposed method achieves a mean average precision (mAP@[0.5:0.95]) of 97.62%, outperforming several prominent deep learning detection methods such as YOLOv8n and MobileNet-SSD. The proposed system provides robust technical support for precision disease prevention and control, intelligent early warning, and sustainable agricultural development, with broad application prospects and significant industrial value.