ACRC-YOLO: An Improved Algorithm for Insulator Defect Detection

To enhance detection accuracy under complex backgrounds and environmental noise, this paper proposes an improved YOLOv11-based algorithm for insulator defect detection. The method introduces a downsampling ADown module to replace part of the traditional Conv layers, reducing spatial redundancy and enhancing semantic feature extraction. A Coordinate Attention (CA) mechanism is embedded to strengthen feature fusion, while a Res_CBAM module—combining CBAM with residual connections—is added to refine feature representation and improve target focus. Experiments on a public dataset show that the model improves [email protected] by $3.9 \%$ and $\mathrm{mAP} \text{@} 0.5: 0.95$ by $2.9 \%$, validating its effectiveness and application potential.