By improving YOLOv5s and SIFT for cartoon characters Detection methods to prevent infringement

Abstract

In recent years, with the popularity of the Internet, a large number of copyright infringement cases continue to emerge. Especially in the cartoon industry, due to the difficulty of replicating original works, infringement incidents are more frequent. Therefore, this article aims to prevent cartoon character infringement and designs and implements a cartoon character detection method based on improved YOLOv5s and SIFT algorithms. In response to the weak recognition ability of YOLOv5s model for small-sized objects, the YOLOv5s algorithm is improved by introducing multi-scale prediction, adding new categories, adjusting parameters, and adjusting weights. The experimental results show that the improved YOLOv5s and SIFT algorithms have improved accuracy, recall, and F1 score, and significantly improved computational efficiency when processing large-scale images.