Traffic Sign Detection Algorithm Based on Improved YOLOv8n

Aiming at the problems of low accuracy and easy to miss detection of small targets in the neural network detection model of traffic signs, an improved target detection algorithm based on YOLOv8n is proposed. The method introduces the MLCA attention mechanism to C2f, enhances the sensory ability, and optimizes the feature extraction. For the problem of weak small target capturing ability appeared in the model, a small target detection layer is added to improve the small target global feature capturing ability. Finally the detector head adopts a dynamic target detection head DyHead (dynamic head) that combines the attention mechanism, which is unified by the triple unification of scale-awareness, spatial-awareness, and task-awareness in order to achieve a stronger feature expression capability. Experiments are conducted on a tt100k dataset as the basis, and the experimental surface shows that the improved YOLOv8 algorithm improves from 78.1% to 87.3% on mAP50.