Yolov8-CS: A Traffic Sign Detection Algorithm Based on Improved Yolov8

In the context of the rapid advancements being made in autonomous driving technology, traffic sign detection has emerged as a pivotal technology to ensure safe driving. However, existing traffic sign detection algorithms are beset with challenges, including difficulties in feature extraction and model bias when identifying small targets and dealing with imbalanced data. To address these issues, this paper proposes an enhanced algorithm, termed YOLOv8-CS (CBAM, SlideLoss), which is based on YOLOv8 and incorporates the CBAM (Convolutional Block Attention Module) attention mechanism to enhance the model's capacity to identify key features. This, in turn, leads to an improvement in the detection accuracy for small traffic signs. Furthermore, the loss function in the YOLOv8 model is enhanced, and SlideLoss is employed to augment the model's capacity to handle imbalanced data, ensuring the model focuses more on challenging cases in data training. The experimental results demonstrate that, in comparison with the original YOLOv8 model, the enhanced YOLOv8-CS model exhibits a 6.1% increase in accuracy, a 4.4% increase in recall rate, and a 5.1% increase in mAP. Furthermore, the enhanced model demonstrates a comparable training speed to the original model, thereby substantiating its viability for practical applications.