Abstract: To address the challenges of small object detection in Unmanned Aerial Vehicle images, including detection difficulty, limited computational resources, and high real-time needs, a lightweight detection model named DroneNet-Lite is proposed. Based on YOLO11, the model includes three key novel modules. First, a P2 small object enhancement layer is added for small object enhancement. Second, a Depthwise Spatial Pyramind Pooling Fast(DWSPPF) module is designed to reduce computational complexity by replacing standard convolutions with depthwise convolutions. Finally, a Channel Fusion Bottleneck (CFB) module is constructed to prevent feature loss du-ring channel changes. The CFB module fuses high-dimensional and low-dimensional features through channel transformation mechanisms. Comparative experiments were conducted on the VisDrone2019-DET dataset between DroneNet-Lite and detection models including YOLOv8, YOLO11, and TPH-YOLOv5, and ablation experiments were performed on the improvement modules of DroneNet-Lite. Comparative experimental results demonstrated that DroneNet-Lite-n (1.9 M parameters) achieves mAP50 improvements of 2.2% and 1.9% over YOLOv8-n and YOLO11-n, respectively. DroneNet-Lite-s (7.0 M parameters) attains 44.6% mAP50, surpassing TPH-YOLOv5-s and other comparable models. DroneNet-Lite-m (16.9 M parameters) reaches 49.6% mAP50, representing a 3.1% improvement over YOLO11-m while reducing parameters by 16%. Ablation experimental results indicate that the P2 small object enhancement layer, DWSPPF module, and CFB module all effectively enhance detection performance. Compared with YOLO11, introducing the P2 small object enhancement layer raises mAP50 by 2.7% and cuts parameters by 25.3%. The DWSPPF module trims 8.4% of parameters while slightly improving detection accuracy. The CFB module adds another 0.2% mAP50 gain. Compared to individual module integration, DroneNet-Lite achieved mAP50 improvements of 1.0%, 2.9%, and 3.5% for P2 small object enhancement layer, DWSPPF module, and CFB module, respectively. In conclusion, DroneNet-Lite significantly enhances small object detection perfor-mance while maintaining lightweight characteristics, providing an efficient detection solution for resource-constrained embedded devices.