GEE-UOD: An Underwater Object Detection Network Based on Global and Edge Information Enhancement

Underwater object detection is crucial for marine exploration but faces challenges like low image quality, severe object overlap, and detection of small, diverse targets. This paper presents a novel model, Global and Edge Information Enhancement Underwater Object Detection model (GEE-UOD), designed to address these issues and improve detection accuracy. The Global and Edge Information Enhancement Module (GEI-EM) extracts global and edge features via multi-branch convolution, improving feature robustness. Additionally, the integration of GLSA, ELA attention mechanisms, and the BiFPN network boosts multi-scale feature extraction and fusion. To balance accuracy and efficiency, a shared convolution detection head is introduced, reducing model parameters and computational cost. Experimental results on the RUOD dataset show that GEE-UOD achieves an mAP of 86.9%, demonstrating its superior performance and potential for advancing underwater object detection.