Research on Moving Object Detection of Animated Characters
Jinyan Liu, De Li · Procedia Computer Science · 2022
In recent years, excellent 3D animation works continue to appear, 3D animation copyright protection has begun to be paid attention to. At present, most of the copied animations copy the movements and movements of characters. In order to carry out copyright authentication and plagiarism detection for such plagiarism, it is necessary to carry out target detection for animation characters first. In this paper, Yolov5, a target detection method based on deep learning, is used for target detection of characters in 3D animation. CSPDarknet53 feature extraction network is used in Yolov5, Focus structure is used as the benchmark network, and FPN+PAN structure is added, which not only improves the accuracy and speed, but also greatly improves the detection of small targets. In this paper, SENET attention mechanism module is added in Yolov5, which reduces the speed of target detection and improves the accuracy of target detection, and is more suitable for the applicability of moving target detection of animated characters.