Small object detection is essential for applications such as UAV monitoring; however, it remains challenging due to the limited size of targets, complex backgrounds, and insufficient feature details. To overcome these challenges, we propose EIL-YOLO, a novel model built upon YOLOv8, which balances accuracy and efficiency through optimized performance and lightweight design. The model introduces three key innovations: (1) a lightweight attention module to enhance small target feature extraction; (2) an optimized neck structure with a scale fusion branch for effective integration of shallow and deep features; (3) a lightweight convolution module for efficient multi-scale feature processing, improving feature representation and network generalization. We evaluated the proposed model on publicly available small-object datasets. The model achieved mAP50 scores of 45.4% on the VisDrone2019 dataset, corresponding to improvements of 20.1%, respectively, over YOLOv8s, while achieving a real-time inference speed 2.7 times faster than YOLOv8s. Furthermore, the model has a parameter size of only 4.25 million, which constitutes 38.2% of the parameter sizes of YOLOv8s. These results demonstrate that the proposed model not only achieves superior detection accuracy in small-object detection tasks but also exhibits significant lightweight characteristics, underscoring its practical application value.