<p>Logistics visual inspection involves several practical difficulties at once, including subtle parcel appearance defects, dense small barcodes under cluttered backgrounds, and strict runtime-efficiency constraints in practical deployment settings. To handle these conditions, LQ-YOLO11 is built on YOLO11n as a lightweight dual-gated detector for logistics inspection. The model combines shallow feature refinement with deep semantic enhancement, so that detection quality can be improved without sacrificing runtime efficiency. In the shallow backbone, Gated Shuffle Partial Convolution (GSPConv) reduces spatial redundancy through channel splitting and shuffle interaction while preserving fine local structures at relatively low cost. In deeper layers, Grouped Shuffle Thresholded CBAM (GST-CBAM) performs grouped semantic recalibration with hybrid gating, which helps suppress background interference and strengthen task-relevant responses. Experiments on Parcel3D and BarBeR indicate that the proposed design maintains a favorable balance between accuracy and efficiency. Relative to YOLO11n, the parameter count decreases from 2.6 to 2.2 M and the computational cost drops from 6.5 to 5.2 GFLOPs, while inference speed increases from 308 to 315 FPS. The detector achieves 0.984 mAP@0.5:0.95 on Parcel3D and 0.921/0.927 on BarBeR at input resolutions of 640 and 960, respectively. These results suggest that LQ-YOLO11 provides favorable deployment-oriented efficiency and accuracy for real-time logistics visual quality inspection under the current experimental setting.</p>