Small traffic sign detection based on improved YOLOv11n

With advances in intelligent transportation and autonomous driving, accurate traffic sign detection is vital for safety, yet detecting small signs in complex scenes remains difficult. This paper presents SRC-YOLO, a lightweight model based on YOLOv11, that improves small object detection while ensuring efficient inference. The P5 detection layer is removed to reduce interference and parameters. A C3k2_SWC module in the backbone enhances multi-scale feature extraction using combined kernel convolutions. V7DownSampling preserves small object details during reduction, and a C3k2_RFAConv module in the neck strengthens spatial modeling with receptive field attention. Experiments on CCTSDB 2021 and TT100K show SRC-YOLO achieves superior accuracy, efficiency, and robustness, especially for small signs.