Spatial-aware lightweight network for real-time tea disease detection: A coordinate attention-enhanced YOLOv8n approach with path-decoupling strategy

Abstract

The intelligent identification of tea diseases is crucial for ensuring tea quality and reducing economic losses in the tea industry. However, the deployment of deep learning models on edge devices remains challenging due to the conflict between detection accuracy and computational overhead. To address this, we propose CA-YOLOv8n, a lightweight object detection model tailored for tea disease diagnosis. Specifically, we introduce a Path-Decoupling strategy to streamline the network structure and integrate the Coordinate Attention (CA) mechanism to enhance the model's spatial awareness of subtle pathological features. Experimental results demonstrate that the proposed model achieves a mean Average Precision (mAP@50) of 98.89% while reducing the parameter count by 32.6% and FLOPs by 24.1% compared to the baseline YOLOv8n. The model was integrated into a diagnostic platform with an automated reporting interface, demonstrating that real-time tea disease identification is feasible on commodity CPU hardware in resource-constrained agricultural environments.