YOLO-SK is a lightweight multiscale object detection algorithm that improves upon the YOLOv5 model by introducing a weighted dense feature fusion network and SK attention prediction head. The weighted dense feature fusion network dynamically fuses features at different scales using learnable parameters and cross-layer fusion, enhancing the richness of effective information in the fused feature maps. The SK attention mechanism broadens the model's receptive field and sharpens its focus on target characteristics, enabling more accurate predictions. The SIoU loss function and Ghost Conv are also implemented to improve accuracy and reduce model complexity. Extensive testing on the PASCAL VOC 2007 and 2012 datasets showed that YOLO-SK achieved a 2.6% increase in mAP@.5 and a 4.8% increase in mAP@.5:.95 compared to the baseline model, while maintaining the same level of model complexity. The model's improvements include a new weighted dense feature fusion network (WD_FPN-PAN), SK attention prediction head (SK_PH), SIoU loss function, and Ghost Conv. These enhancements allow YOLO-SK to detect objects of various scales with high accuracy while keeping the model lightweight and efficient. The model's performance was validated through experiments, showing improved detection accuracy and convergence performance compared to other models. YOLO-SK is suitable for deployment on low-performance devices due to its lightweight design and efficient computation.YOLO-SK is a lightweight multiscale object detection algorithm that improves upon the YOLOv5 model by introducing a weighted dense feature fusion network and SK attention prediction head. The weighted dense feature fusion network dynamically fuses features at different scales using learnable parameters and cross-layer fusion, enhancing the richness of effective information in the fused feature maps. The SK attention mechanism broadens the model's receptive field and sharpens its focus on target characteristics, enabling more accurate predictions. The SIoU loss function and Ghost Conv are also implemented to improve accuracy and reduce model complexity. Extensive testing on the PASCAL VOC 2007 and 2012 datasets showed that YOLO-SK achieved a 2.6% increase in mAP@.5 and a 4.8% increase in mAP@.5:.95 compared to the baseline model, while maintaining the same level of model complexity. The model's improvements include a new weighted dense feature fusion network (WD_FPN-PAN), SK attention prediction head (SK_PH), SIoU loss function, and Ghost Conv. These enhancements allow YOLO-SK to detect objects of various scales with high accuracy while keeping the model lightweight and efficient. The model's performance was validated through experiments, showing improved detection accuracy and convergence performance compared to other models. YOLO-SK is suitable for deployment on low-performance devices due to its lightweight design and efficient computation.