Research on rubbish detection in scenic spots based on YOLOv8 algorithm

A lightweight YOLOv8n litter detection model is proposed for the study of the existing scenic litter problem. The LSKA module is introduced into the SPPF, andthe large kernel separation convolution operation is used to weight different feature channels according to their importance, which improves the ability of the model to extract feature information. The convolution of the backbone network is replaced with DWConv convolution, which effectively maintains the accuracy and reduces the number of parameters for easy deployment to mobile devices. Experiments show that the LD-YOLOv8n model achieves an average accuracy of 89.7% and an improvement of 1.7% compared to theYOLOv8n model, while the number of parameters and the number of floating-point operations per second are reduced by 4% and0.7G, respectively.