An Improved Lightweight YOLO Model for Real-Time Blueberry Growth Stage Detection to Facilitate Smart Irrigation

Accurate detection of power transmission towers in remote sensing images is crucial for grid inspection and infrastructure monitoring. However, existing detectors often suffer from high computational costs or limited accuracy in complex backgrounds. This paper presents a lightweight detector specifically designed for power transmission tower detection. We evaluate two efficient model architectures: YOLOv8 integrated with a CBAM attention mechanism (YOLOv8-CBAM) and a real-time detector RT-DETR enhanced with Swin Transformer backbone (RT-DETR+Swin). Experimental results on remote sensing imagery demonstrate that YOLOv8-CBAM achieves 88.2% accuracy, while RT-DETR+Swin yields a superior 91.8% accuracy. To facilitate practical deployment, we develop an interactive web-based frontend using Streamlit, enabling real-time visualization and inference. The proposed system balances detection performance and lightweight design, making it suitable for resource-constrained remote sensing applications.

Keywords: Power transmission tower detection; remote sensing image; lightweight detector; YOLOv8-CBAM; RT-DETR with Swin Transformer; Streamlit.