EdgeEnhance-YOLO: A Lightweight Small Object Detection Model With Multidimensional Edge Enhancement

Small object detection faces multiple challenges, including high-frequency information degradation, complex back ground interference, and low target resolution, which frequently lead to missed and false detections. To address these issues, this paper proposes an end-to-end multidimensional edge-enhanced detection network, named EE-YOLO, which exhibits significant technical innovations in structural design and feature fusion mechanisms. Firstly, an Edge-Enhanced Convolution module (EEConv) is designed with a dual-branch parallel architecture, enabling the synchronous extraction of spatial information and fine-grained edge details in shallow network layers, thereby effectively mitigat ing high-frequency information loss. Secondly, an Edge-Enhanced Multi-Scale Feature Fusion Framework (EEMFF) is developed to overcome the limitations of conventional feature pyramid structures. Through a direct edge injection mechanism, high-resolution shallow edge features are fused across layers into deep semantic representations, while a four-scale fusion module provides enriched contextual information to enhance small-object detection capability globally. Finally, to balance detection accuracy and computational efficiency under resource-constrained conditions, a Detail-Enhance Lightweight Shared Convolution Detection Head (DELSCD) is constructed, which maintains high-precision detection while sig nificantly reducing model parameters. Extensive experiments con ducted on the VisDrone2019, TinyPerson and RS-STOD datasets demonstrate that EE-YOLO, with fewer than 10M parameters, surpasses existing publicly available models, achieving 51.0% and 33.5% mAP(50) and mAP(50-95) on VisDrone2019, 30.9% and 12.0% on TinyPerson, and 83.2% and 58.8% on RS-STOD, respectively. The research results achieve both technical breakthroughs in lightweight small object detection and engineering advances for real-world visual perception tasks, offering an innovative solution that effectively balances detection accuracy and deployment efficiency. The proposed EE-YOLO framework demonstrates strong scalability and applicability across a wide range of practical scenarios, including intelligent transportation monitoring, public safety monitoring, and industrial inspection environments.