Pedestrian detection algorithm for scenic spots based on improved YOLOv8

Aiming at the current pedestrian target detection in the scenic area, there are issues like inadequate detection precision and the challenge of identifying small-sized pedestrians, an improved YOLOv8 scenic area pedestrian detection algorithm is proposed. Firstly, the C2fNEW convolution module is designed to expand the detection field of view, which improves the detection ability of the model; secondly, the BiFormer attention mechanism and the replacement of the DyHead dynamic detection head are used to improve the detection of the pedestrian features as well as the expression ability of the detection head; lastly, the improved channel enhancement module is designed to enhance the model's ability of extracting the features of the channel, which effectively improves the model's ability to detect pedestrians in scenic spots. Through the analysis of experimental results, the improved YOLOv8 algorithm performs well on the pedestrian label data set in PASCAL VOC2007+2012, with an increase of 3.1% at [email protected] and 2.9% at [email protected]:0.95, which represents a notable enhancement over the initial algorithm. It can be better applied in pedestrian detection tasks in scenic spots.