Answers are grounded in this paper and cite their sources.
To overcome the bottleneck of missed detection, false detection and low recognition accuracy in complex environment, This paper proposes HGR-YOLO, a specialized architecture for transmission tower defect identification, distinguished by three strategic enhancements: High-Order Interactions Network (HorNet) is designed to perform efficient, scalable and isovariable high-order spatial interactions through recursive design and gated convolution, and enhance feature extraction capabilities to achieve the goal of accurate defect identification. The Ghost Convolution (GhostConv) module is designed to synthesis a rich ensemble of feature representations based on a set of intrinsic feature mappings, which can improve accuracy and reduce computational overhead. Finally, a Re-parameterization Vision Transform block (RepViTblock) is proposed to unwrap the channel mixer and the token mixer through structural reparameterization technology to enhance model learning during training. The enhanced model achieves a precision of 93.5%, a recall of 87.9%, and an mAP@50 of 92.7%. Relative to the baseline, the precision, recall and mAP@50 represent substantial gains of 1.6%, 7.5%, and 5%, respectively. Such improvements underscore the model’s robust detection capabilities, positioning it favorably against prevailing mainstream methods.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: