A Lightweight Object Detection Model Based on Improved Yolov8

Aiming at the complex structure of YOLOv8 target detection algorithm and the large amount of algorithm parameters and calculations, it's challenging to deploy in terminals. We propose a lightweight detection model, LW-YOLOV8, based on an enhanced version of YOLOv8. Firstly, the two-way feature pyramid network of the neck is replaced by a weighted Bidirectional characteristic pyramid structure (BIFPN) and combined with the FasterNet network idea, FasterBifpn is designed. Secondly, a lightweight detection head with shared parameters based on attention mechanism is designed, which significantly reduces the number of parameters while maintaining detection accuracy. In the loss function, the Inner-SIOU is introduced in combination with the Inner-IoU idea, which not only converges faster, but also further optimizes the accuracy of the detection frame. The experimental results show that, compared with the original model, the accuracy of the LW-YOLOv8 model ([email protected]) is only reduced by 0.05 %, while the number of parameters is reduced by 48.3 %, and the amount of computation is reduced by 32 %, achieving a lightweight effect.