YOLOv11-PLight: An Efficient Object Detection Model Based on Lightweight Data Processing

Achieving high-precision object detection on resource-constrained devices remains a significant challenge, especially when dealing with multi-scale objects, small targets, complex backgrounds, and occluded scenes, where balancing detection accuracy and inference efficiency is difficult. To address this, this paper proposes an efficient lightweight object detection model, YOLOv11-PLight, to overcome the performance bottlenecks faced by existing methods in practical deployments. The model integrates three key modules: the PConv Backbone network enhances feature robustness in occluded and missing regions, the C3k2SC module improves multi-scale feature perception, and the ECSA-Net module achieves cross-layer spatial and channel attention fusion. Experiments on the COCO and Cityscapes datasets show that YOLOv11-PLight achieves significant improvements across multiple performance metrics, with accuracy gains ranging from 3.5% to 6.8% and inference efficiency improvements of 8.4% to 15.2%. Through innovative module design, YOLOv11-PLight not only achieves a better balance between accuracy and computational complexity but also optimizes model performance on edge devices, providing a more practical solution for object detection in edge computing environments, significantly reducing computational overhead while maintaining detection accuracy.