Research on Multi-Target Detection Method Based on YOLOv8 in Multi-Scenes

Aiming at the current mainstream target detection algorithms in a variety of different scenarios for multi-target detection still exists in the target scale difference, susceptible to environmental interference, and poor real - time performance, this paper puts forward a multi-target detection method based on the YOLOv8 model improvement. This algorithm preprocesses the contrast and resolution of the image before the backbone network, enhances the target features in the image, and uses the attention mechanism to increase the attention to the target features. This paper will also use the combined loss function to comprehensively consider multiple performance indicators, and make the model achieve better overall performance by weighing different loss items, so as to more accurately identify the target. In the process of this experiment, the VisDrone2019 dataset will be used to carry out ablation experiments, and the experimental results will be evaluated in relation to other algorithms. The experimental findings demonstrate that the method proposed in this paper leads to substantial enhancements in relevant performance metrics and effectively addresses the accuracy issue in multi-target detection.