<p>Existing deep learning models have demonstrated remarkable performance in remote sensing image object detection; however, their substantial computational complexity and high resource consumption hinder deployment on edge devices with limited hardware capabilities. To address this challenge, we propose a lightweight and effective YOLO-based model (LE-YOLO) for fast and accurate object detection in remote sensing images. Specifically, a new backbone network is introduced by replacing the C2f module in the Yolov8n backbone with the DualHet module, which strengthens feature representation while reducing complexity. Then, the lightweight hybrid convolutional–transformer detector (HCT-Det) is developed to effectively model global contextual dependencies and improve the discrimination of salient objects within complex remote sensing scenarios. Furthermore, the Fast-IoU equipped with a dynamic non-monotonic focusing mechanism is adopted to efficiently refine medium-quality anchor boxes, further enabling more accurate localization and detection performance. Experimental results on the AI-TOD and Visdrone datasets show that our LE-YOLO achieves mAP50 of 45.8% and 39.3% respectively, with only 2.14 MB parameters and 5.7 GFLOPs, striking a balance between detection accuracy and computational efficiency. Finally, the proposed LE-YOLO is successfully deployed on an embedded system composed of a Rockchip RK3588 and a low-cost embedded image sensor, enabling the real-time object detection in remote sensing applications.</p>