Improved YOLOv8 Lightweight Floating Bottles Target Detection Algorithm

Owing to the phenomena of missed and false detections of floating bottles in river channels, and simultaneously, the large number of model parameters and high computational complexity, which is not conducive to end-to-end industrial deployment, a lightweight floating bottle target detection algorithm with improved YOLOv8 was proposed. Firstly, the reconstructed C2fFaster module in the backbone part can effectively reduce the computational amount and number of parameters of the model. Secondly, with the aim of enabling the model to focus more closely on small targets and consequently reducing the rate of missed detections, the Triplet attention module was introduced. Finally, in the feature fusion part, the improved BiFPN structure was adopted, which fused the P2 layer to obtain more feature information of small targets. And the reconstructed C2fFaster module was used to improve detection accuracy and speed. Compared with YOLOv8n, the accuracy is increased by 4.9%. The total number of parameters and the number of floating-point operations are decreased by 48.17% and 31.71% respectively. The improved algorithm effectively balances model lightweight and performance and provides a new idea for end-to-end industrial deployment.