Road Sign Detection Based on Yolov8
Yao Guo, Yuansong Li · 2025
Addressing the issues of scale diversity and poor detection of small, distant objects in traffic sign recognition, this research introduces an advanced algorithm for traffic sign detection, which is an enhancement of YOLOv8. This new approach embeds a Channel Attention (CA) mechanism into the C2f module of YOLOv8's foundational network to boost the model's ability to concentrate on key features. Additionally, a Space-to-Depth Convolution (SPDConv) module is integrated to reduce the loss of detailed information and to strengthen the network's feature extraction across different scales. The algorithm's performance was tested on the CCTSDB dataset, with the results showing a detection accuracy of 97.7%, which is a significant increase of 7.4% compared to the previous version. The upgraded algorithm shows higher precision and stability, proving its efficacy in real-world scenarios.